934
THE RED SECTION
nature publishing group
The Quality of Care Provided to Patients With Varices in the Department of Veterans Affairs Paula M. Buchanan, PhD1, Jennifer R. Kramer, PhD2, Hashem B. El-Serag, MD, MPH2,3, Steven M. Asch, MD, MPH4, Youssef Assioun, MD5, Bruce R. Bacon, MD5 and Fasiha Kanwal, MD, MSHS2,3 OBJECTIVES:
Practice guidelines define the criteria and standards of care in patients with cirrhosis and varices. However, the extent to which the patients receive recommended care is largely unknown. We evaluated the quality of varices related care and factors associated with receipt of such care.
METHODS:
We conducted a retrospective cohort study of 550 patients with cirrhosis who sought care at three VA facilities between 2000 and 2007. Using administrative and clinical data, we assessed quality of varices care as measured by eight explicit Delphi panel-derived quality indicators. We also conducted a structured implicit review of patients’ medical records to explore the role of patients’ refusal, receipt of care outside the VA, or justifiable exclusions to certain care processes as explanations for non-adherence to the quality indicators.
RESULTS:
Quality scores (max. 100%) varied across individual indicators, ranging from 24.3% for upper endoscopy for varices screening to 72.4% for secondary prophylaxis for variceal bleeding. Justifiable exclusions to indicated care documented in charts were common for primary prophylaxis in patients with varices; receipt of endoscopy; and endoscopic treatment in patients with active bleeding. In contrast, significant shortfalls remained in the receipt of screening endoscopy, use of beta-blockers (in the absence of varices), and use of somatostatin analogs, antibiotics, and secondary prophylaxis in patients with variceal bleeding. Younger patients ( < 60 vs. > 60 year, odds ratio (OR) = 1.29, 95% confidence interval (CI) 1.01–1.68), those who saw a gastroenterologist (OR = 1.55, 95% CI = 1.09– 2.21), or those who were seen in the facility with academic affiliation (OR = 1.26, 95% CI = 1.01– 1.58) received higher quality care.
CONCLUSIONS:
Health-care quality, measured according to whether patients received recommended varices-related care, was suboptimal in this health-care setting. Care that included gastroenterologists was associated with high quality.
Am J Gastroenterol 2014; 109:934–940; doi:10.1038/ajg.2013.487
Gastroesophageal varices are present in almost half of the patients presenting with cirrhosis at the time of initial diagnosis (1,2). Variceal bleeding occurs in 25–35% of patients with cirrhosis (3). Up to 15% of initial bleeding episodes are fatal (4), and 70% of survivors may have recurrent bleeding (5). Despite significant developments in the management of varices, mortality from bleeding varices remains high (4). Effective and timely care can improve outcomes in patients with varices and variceal bleeding. For example, clinical studies demonstrate that both beta-blockers and endoscopic variceal ligation (EVL) are effective in preventing first episode
of variceal bleeding (6–9). Results of controlled trials show that EVL is superior to sclerotherapy for control of active bleeding, and that this superiority is consistent across all key outcomes (10). Similarly, data show that use of somotostatin analogs reduces the risk of initial treatment failure and blood transfusions in patients with active variceal bleeding (11,12). Randomized controlled studies also demonstrate the benefit of using antibiotic prophylaxis in patients with variceal bleeding (13). Despite the available evidence, existing data indicate the presence of systematic deficiencies in the care that is provided to patients
1
Center for Outcomes Research, Saint Louis University, Saint Louis, Missouri, USA; 2Michael E. DeBakey Veterans Affairs Medical Center, Houston, Texas, USA; 3Division of Gastroenterology, Department of Medicine, Baylor College of Medicine, Houston, Texas, USA; 4Department of Medicine and Health Services Research, Palo Alto VA Healthcare System, Palo Alto, California, USA; 5Department of Medicine, Saint Louis University, Saint Louis, Missouri, USA. Correspondence: Fasiha Kanwal, MD, MSHS, Michael E. DeBakey Veterans Affairs Medical Center, 2002 Holcombe Boulevard (152), VA, BCM, Houston, Texas 77030, USA. E-mail: [email protected] or [email protected] The American Journal of GASTROENTEROLOGY
VOLUME 109 | JULY 2014 www.amjgastro.com
THE RED SECTION
with varices. For example, Wilbur and Sidhu (14) found that 94% of patients with variceal bleeding had not received any primary prophylaxis. Recent data also suggest wide variations in the treatment of active bleeding across treating providers. For example, in a multicenter study from the United States including 700 patients, endoscopic intervention included banding in 41%, sclerotherapy in 36%, and both modalities in 6% of patients (15). Similarly, Singh et al. (16) found that follow-up endoscopy for secondary prophylaxis was arranged for only 65% of the patients after endoscopic treatment of the initial bleeding episode. These examples highlight that the quality of care in the management of varices may be suboptimal. However, these studies have focused on a single process each. No study has examined quality of care across the sequential processes of care—from diagnosis to treatment and preventive care—in the management of patients with varices. We recently developed and published an explicit set of evidence-based quality indicators (QIs) in cirrhosis using the modified Delphi panel approach (17). In a previous report, we used the ascite-specific QIs to assess the quality of ascitesrelated care (18). We now extend our evaluation to assess the quality of varices care in a large sample of patients with cirrhosis who sought care at three Veterans Affairs (VA) medical centers between 2000 and 2007. We also examined the relationship between patient factors and quality of care, and explored the role of justifiable exclusions, patients’ refusal, and receipt of care outside the VA health-care system as possible explanations for receipt of lower quality care.
Methods
Using an administrative and clinical database, we determined the adherence to an explicit set of varices-specific QIs in a cohort of patients with cirrhosis. We then conducted a structured implicit review of patients’ medical charts to explain findings from the explicit analysis. Quality indicators. Quality of variceal bleeding care was measured using eight specific QIs. These QIs were derived from a larger set of QIs measuring care provided to patients with cirrhosis, as described before (17). The original set included 18 QIs in the variceal bleeding care domain. Although all were rated as appropriate indicators of quality, the amount and quality of evidence as well as the size and precision of the treatment effect to support individual QIs varied. For this analysis, we included only those varices QIs that were supported by evidence from randomized trials or meta-analyses (Levels A and B per GRADE method) (19). The one exception was the “receipt of upper endoscopy” to screen for varices because it has widespread expert support from the guidelines. Table 1 lists the QIs that we examined in this study. They included three QIs measuring pre-bleeding care (receipt of upper endoscopy to screen for varices, receipt of primary prophylaxis with beta-blockers or EVL, and not receiving beta-blockers in the absence of documented varices); four QIs measuring variceal bleeding care (receipt of somatostatin analogs, antibiotics, upper © 2014 by the American College of Gastroenterology
endoscopy, and either EVL or sclerotherapy); and one QI measuring post-bleeding care (secondary prophylaxis with beta-blockers and EVL or sclerotherapy). Table 1 also shows the numerator, denominator, and time frame for each measure during which an indicated process was required to occur. For example, patients were classified as having met the “received an upper endoscopy to screen for varices” QI if they received an upper endoscopy within 12 months of cirrhosis diagnosis. Data source and study sample. We used a database derived from the VA electronic medical record system that included information for patients who sought care at 3 VA Medical Centers and their 15 affiliated clinics in VISN-15 of the Midwest United States. The database includes separate files for patient demographics, outpatient clinic and inpatient utilization, vital signs, pharmacy utilization, and laboratory data. We defined cirrhosis patients as adults with at least two ICD-9 codes for cirrhosis (571.2, 571.5, and 571.6) or at least one ICD-9 code for cirrhosis in combination with either a code for complications of cirrhosis (456.0, 456.1, 456.20, 456.21, 572.2, 572.3, 572.4, 572.8, and 789.5) or aspartate aminotransferase: platelet ratio > 2 between January 2000 and December 2007 (20). Patients had to be eligible for at least one QI to be included in the analysis. Chart validation of database-derived variables. For each QI, we developed specifications for measurement based on a combination of sources (ICD-9 codes and CPT codes, etc.) in the database. We estimated the validity of database-derived definitions of the key variables (such as varices, variceal bleeding, and upper endoscopy) by comparing them with diagnoses derived from a structured and detailed review of the patients’ medical records (Appendix Table A1). We selected 300 patients from the database for the chart validation. A trained clinician abstracted the medical record using a standardized, detailed abstraction form to determine the presence or absence of the key conditions listed above. The abstractor was blinded to the database coding of these conditions. A 10% random sample was additionally reviewed by a senior clinician. The study team discussed and resolved any discrepancies by consensus.
Statistical methods
Chart review validation. We calculated the positive predictive value (PPV), negative predictive value (NPV), and kappa coefficient for database-derived definitions in correctly identifying varices, variceal bleeding, and upper endoscopy in the medical records. PPV is the probability that the condition recorded in administrative data is actually present in the medical record, our gold standard, and NPV is the probability that the condition classified as absent based on administrative data is actually absent according to the medical record. Estimating rates of QI adherence. For each subject, we determined eligibility for the process specified in each QI (i.e., included in the denominator). We then determined the receipt of The American Journal of GASTROENTEROLOGY
935
936
THE RED SECTION
Table 1. Rates of quality indicators, defined as percentage of cirrhosis patients who received the care indicated by each indicator (numerator) among those who were eligible for the indicator (denominator) QI
Rate % (95% confidence interval) (denominator)
Quality indicators
Pre-bleeding domain 1
Among patients with cirrhosis and no documented history of variceal bleedinga, % who received an upper endoscopy within 12 months of cirrhosis diagnosis
24.3 (20.2-28.3) (441)
2
Among patients with cirrhosis and no documented history of variceal bleedingb, % who did not receive non-selective beta-blockers for primary variceal prophylaxis
72.4 (68.0–76.9) (388)
3
Among patients with cirrhosis, no documented history of variceal bleedingc, but with varices on endoscopy, % who received either non-selective beta-blockers or esophageal variceal ligation (EVL) within 1 month of varices diagnosis (including non-selective beta-blockers before varices diagnosis)
60.4 (50.2–70.7) (91)
4
Among patients with cirrhosis who have variceal bleeding on or after cirrhosis index date, % who receive somatostatin or analogs (somatostatin, octreotide, terlipressin) within the index hospitalization
59.3 (50.1–68.5) (113)
5
Among patients with cirrhosis who have variceal bleeding on or after cirrhosis index date, % who receive an upper endoscopy within 24 hours of presentation
74.8 (66.9–82.7) (119)
6
Among patients with cirrhosisd and variceal bleeding, % who receive EVL or sclerotherapy within 1 week of varcieal bleeding date
63.5 (51.3–75.7) (63)
7
Among patients with cirrhosis who have variceal bleeding, % who receive antibiotic prophylaxis during hospitalization
37.7 (28.7–46.8) (114)
Bleeding-related domain
Secondary prevention domain 8
Among patients with cirrhosis who have variceal bleeding who survived for 3 months post bleeding, % who received either nonselective beta-blockers and/or EVL or sclerotherapy within 3 months of varcieal bleeding date
72.4 (62.6–82.0) (87)
a
No bleeding 1 year before cirrhosis index date through endoscopy date. No bleeding for 1 year before cirrhosis index date and no varices from 1 year before to 1 year after cirrhosis index date. c No bleeding until 1 day after the endoscopy date. d Removed patients with gastric varices codes from the denominator. b
recommended care within the specified time frame. We calculated the rate of each QI as the percentage of subjects who received the recommended care from among those subjects who were eligible for the QI. Implicit chart review. Using the chart validation sample described above, we conducted a structured review of patients’ medical records to explore justifiable explanations for non-adherence to the indicated care processes that were not covered by the QI specifications above. These included, but were not restricted to, receipt of care outside the VA, being too sick to receive indicated care, or patients’ refusal. For each QI, we only reviewed cases that were eligible for, but did not receive the care as indicated by the QI. Predictors of quality care. We used multivariable analysis to identify factors associated with receipt of varices-related care. The dichotomous dependent variable was receiving the recommended care for each QI. Each patient had a row of data for each QI for which they were eligible. As our data were hierarchically structured— with QIs clustered within patients—we constructed multilevel generalized linear mixed models (GLIMMIX in SAS (SAS, Cary, NC)) with random effects for the patients to determine predictors The American Journal of GASTROENTEROLOGY
of varices care. The independent variables included demographic factors (age, race), liver disease severity-related characteristics (model for end-stage liver disease or MELD score (21), presence of ascites), and medical comorbidities (Deyo index) (22). Table 2 provides the full list, operational definitions, and categorization of these patient factors. We defined patients as having seen a specialist if she/he had a visit (either inpatient or outpatient, based on the QI) with a gastroenterologist during the observation period (i.e., time frame) for any of the indicated QIs. We assigned each patient a facility where the patient was seen most frequently. We created a variable indicating the academic affiliation of the facilities. We reported our results as odds ratios (ORs) with 95% confidence intervals (CIs), where ORs are the odds of receiving a recommended QI for which a patient was eligible. We considered P < 0.05 as statistically significant. We used SAS (version 9.2) for all statistical analyses (SAS) (23).
Results
Demographics and clinical characteristics. We identified 550 cirrhosis patients who were eligible for at least one variceal bleeding VOLUME 109 | JULY 2014 www.amjgastro.com
THE RED SECTION
Table 2. Characteristics of 550 cirrhosis patients who were eligible for at least one variceal bleeding care quality indicator Variable
Distribution
Demographics Age in years (mean, s.d.)
54.1 (9.4)
Race (N, %) Caucasian
365 (66.4)
African American
87 (15.8)
Unknown
98 (17.8)
Gender (N, %) Male
544 (98.1)
Female
6 (1.1)
Clinical characteristics Etiology of cirrhosis (N, %) Hepatitis C
254 (46.2)
Alcohol
181 (32.9)
Hepatitis B
Results of chart validation. The PPV, NPV, and kappa for our database definition of varices were 89% (95% CI = 74–96%), 90% (84–94%), and 69% (56–81%), respectively. Similarly, the definition of an upper endoscopy had a PPV, NPV, and kappa of 90% (95% CI = 81–95%), 96% (95% CI = 90–98%), and 86% (95% CI = 79–94%), respectively. The PPV for variceal bleeding was 70% (95% CI = 47–86%), whereas the NPV and kappa were 97% (94–99%) and 71% (95% CI = 55–88%), respectively (Appendix Table A1). Rates of QIs. The rate of meeting QIs varied (Table 1). For example, whereas 59.3% (95% CI = 50.1–68.5%) of patients received somatostatin analogs at the time of variceal bleeding, only onethird (37.7%, 95% CI = 28.7–46.8%) received antibiotic prophylaxis during that episode. Similarly, 24.3% (95% CI = 20.2–28.3%) of patients had an upper endoscopy during the first year after their cirrhosis diagnosis. However, 60.4% (95% CI = 50.2–70.7%) of the patients with varices received either beta-blockers or EVL for primary prophylaxis of variceal bleeding. Overall, fewer than 16% (95% CI = 13–19%) patients received all indicated care.
2 (0.4)
Others
113 (20.6)
Complications of cirrhosis (N, %) Ascites
210 (38.2)
Variceal bleeding
123 (22.4)
Hepatic encephalopathy
117 (21.3)
Baseline MELD score (mean, s.d.)a
12.6 (5.3)
Comorbidity index (mean, s.d.)
2.2 (0.9)
Health-care utilization Outpatient visit days per yearb (median, IQR) b
21.4 (12.4–33.2)
At least one hospitalization (N, %)
369 (67.1)
Seen by a gastroenterologist (N, %)
420 (76.4)
IQR=interquartile range. a Data to allow MELD score calculation were available for 502 (91%) of our sample. Baseline MELD is the score closest to the cirrhosis index date. b Outpatient visits or hospitalization occurring after cirrhosis index date.
QI (Table 2). The mean age of our study subjects was 54 (s.d. = 9.4) years, almost all were male, and majority were white (66.4%). Cirrhosis was related to chronic viral hepatitis C and alcohol in 46.2 and 32.9% of patients, respectively. Over a third (38.2%) of the patients had ascites, 22.4% had variceal bleeding, and 21.3% had hepatic encephalopathy. The mean MELD score and Deyo index were 12.6 (s.d. = 5.3) and 2.2 (s.d. = 0.9), respectively. Patients were followed for a mean of 3.3 (s.d. = 1.8) years. The majority of the patients (67.1%) had at least one hospitalization after their cirrhosis diagnosis with a median of 21.4 (IQR = 12.4–33.3) outpatient visits per year. Approximately 76% of patients were seen by gastroenterologists during the time frame specified by at least one of the QIs. © 2014 by the American College of Gastroenterology
Reasons for non-adherence to QIs based on chart review. As shown in Table 3, the proportion of patients with possible justifiable exceptions (i.e., reasons for not receiving care) varied across the QIs. Possible exceptions to indicated care documented in charts were particularly common for “receipt of primary prophylaxis”, “receipt of upper endoscopy in patients with variceal bleeding”, and “receipt of EVL or sclerotherapy for variceal bleeding” QIs. For “receipt of primary prophylaxis” QI, only four patients in our chart review subsample were eligible for but did not meet the QI. Of these, two had small varices, one had low blood pressure (contraindicating beta-blockers but not EVL), and one had metastatic cancer. Similarly, for “receipt of upper endoscopy in patients with variceal bleeding”, three of the four patients received upper endoscopy before the start of eligibility time frame (before admission) and one died within a day of presenting with variceal bleeding. Expanding the time frame to include endoscopy procedures that occur within a few days before the eligibility date (admission date in this case) would capture these patients. Six of the eight chart review patients who did not receive EVL or sclerotherapy for variceal bleeding had justifiable exceptions; three bled from either gastric varices or portal hypertensive gastropathy, one did not have any bleeding, one was too sick, and one received recommended care at an outside facility. In contrast, the majority of patients included in our implicit review for other QIs, including “receipt of upper endoscopy to screen for varices”, “not receiving beta-blockers in the absence of documented varices”, “receipt of somatostatin analogs and antibiotics for variceal bleeding”, and “secondary prophylaxis with betablockers and EVL or sclerotherapy”, did not have an explanation for not receiving the recommended care (classified as having “no documented reason” in Table 3). Predictors of quality care. Results of the multilevel regression analysis are presented in Table 4. Younger patients, those who saw The American Journal of GASTROENTEROLOGY
937
938
THE RED SECTION
Table 3. Reasons for not receiving recommended varices care—results of the implicit chart review in 61 patients with cirrhosis Quality indicators (N of cases included in the implicit review for the indicator)
Number (percentage) of patients No documented reason
Patient refusal
Outside care
Too sick
Other reasons
Endoscopy to screen for varices (n=42)
30 (71.4)
2 (4.7)
0 (0)
1 (0.2)
9 (21.4)
Not receiving beta-blockers in the absence of documented varices (n=15)
14 (93.3)
0 (0)
0 (0)
0 (0)
1 (0.07)
Primary prophylaxis for documented varices (n=4)
0 (0)
0 (0)
0 (0)
1 (25)
3 (75.0)
5 (71.4)
0 (0)
0 (0)
0 (0)
2 (28.5)
0 (0)
0 (0)
0 (0)
1 (25.0)
3 (75.0)
Variceal ligation or sclerotherapy in patients with variceal bleeding (n=8)
2 (25.0)
0 (0)
0 (0)
1 (12.5)
5 (62.5)
Antibiotics in patients with variceal bleeding (n=12)
9 (75.0)
0 (0)
0 (0)
1 (0.83)
2 (16.6)
Secondary prophylaxis (n=4)
3 (75.0)
0 (0)
1 (25.0)
0 (0)
0 (0)
Somatostatin analogs in patients with variceal bleeding (n=7) Endoscopy in patients with variceal bleeding (n=4)
The number of patients who were eligible for but did not meet the individual quality indicators (QIs) are listed with each QI.
a gastroenterologist, or those who were seen in the facility with academic affiliation were more likely to meet the QIs for which they are eligible. For example, patients who were younger than 60 years were 29% more likely to meet the QIs for which they are eligible compared with older patients (OR = 1.29, 95% CI = 1.01– 1.68). If the patient saw a gastroenterologist, then their odds of receiving the recommended care were 55% higher (OR = 1.55, 95% CI = 1.09–2.21), whereas those treated at an academic facility had 26% higher odds of receiving recommended care (OR = 1.26, 95% CI = 1.01–1.58). Removing the QIs that were sensitive to the database limitations (with > 50% rate of justifiable exceptions) did not change the direction or magnitude of the results above.
Discussion
Our study has three key findings. First, we found that the quality of varices care, measured according to whether patients received recommended services, varied substantially across the process of care examined. We found that most patients with documented varices received recommended primary prophylaxis with either nonselective beta-blockers or EVL. Similarly, 75% of patients who presented with variceal bleeding received an endoscopy within 24 h of presentation and most (64%) underwent recommended EVL or sclerotherapy for management of variceal bleeding. In contrast, less than a third of the patients underwent an endoscopy to screen for varices within a year of cirrhosis diagnosis and fewer than 40% received antibiotics during admission for variceal bleeding. Care for other processes fell in between the two extremes. Somatostatin analogs were used in about 60% of patients who presented with acute bleeding, and a quarter of patients did not receive recommended secondary prophylaxis against re-bleeding. These deficits persisted even after accounting for possible justified exceptions in our chart review. Second, we found that automated data underestimated the rates for the measures for which justifiable exceptions for failing to The American Journal of GASTROENTEROLOGY
Table 4. Results of multivariable regression analysis showing the association of demographic and clinical factors with receipt of variceal bleeding care in patients with cirrhosis Variable
Odds ratio
95% Confidence interval
1.29
1.01–1.68
Demographics Age in years (reference ≥60) Race (reference white)
—
—
African American
0.74
0.54–1.01
Unknown race
0.87
0.65–1.15
MELD score (reference ≥10)
1.06
0.84–1.34
Presence of ascites (reference no)
1.04
0.83–1.31
—
—
0/1
1.14
0.86–1.50
2
1.08
0.74–1.21
Seen by a gastroenterologist (reference no.)
1.55
1.09–2.21
Seen in academic facility (reference no.)
1.26
1.01–1.58
Clinical characteristics
Deyo index (reference ≥3)
Health-care system factors
The odds ratio indicate the odds of receiving recommended care for which the patient was eligible. Continuous variables are categorized on the basis of their distribution in the data (i.e., percentiles).
provide the recommended care were common. We found that these exceptions occurred frequently and had a large effect on measured performance for three QIs: receipt of primary prophylaxis, receipt of upper endoscopy in patients with variceal bleeding, and receipt of EVL or sclerotherapy for variceal bleeding QIs. We therefore VOLUME 109 | JULY 2014 www.amjgastro.com
THE RED SECTION
would not recommend the application of these measures using automated data unless a chart review or other validated measures were also used to improve accuracy. However, manual chart reviews may be too inefficient and expensive to be disseminated widely, particularly when used for tracking changes in quality care over time. Other alternatives would be the natural language processing of free text notes or creating new standard codes that explain exceptions or patient preferences for treatment decisions as endorsed by the Centers for Medicare and Medicaid Services’ Physician Quality Reporting System. These codes could be associated with clinical encounters or placed in problem lists to allow capture by automated searches. Third, we found that being seen by gastroenterologists was an important determinant of receipt of quality care in patients with (or those at risk for) varices. These results are consistent with the available data that care by specialists within their area of expertise may provide the best chance at improving the process and/or outcomes of care (24–25). As endoscopy is performed mostly by gastroenterologists, we considered the possibility that the effect of seeing a gastroenterologist on the overall quality of varices care may be mediated solely through its influence on measures related to endoscopic care. However, we found a consistently positive association between being seen by gastroenterologists and each measure separately, although some of these associations were stronger than the others (data not shown). Our data suggest that specialist involvement may improve the quality of care delivered to patients with cirrhosis and varices, although future research will be needed to determine whether this effect is only mediated through direct patient care or could be achieved through tele-consultation with primary care providers. Our study has several limitations. We cannot make causal inferences between QIs and the predictors, given the observational retrospective nature of the study design. Similarly, as the database lacked relevant information, we could not identify specific characteristics of specialists or of the academic setting that explained the respective associations with better care. We also could not assess certain aspects of care such as beta-blocker dose titration and duration, number of bands placed during EVL, patient education, or counseling regarding alcohol use. As some of these parameters (such as dose titration and number of bands placed) are tailored to a given patient’s clinical condition, the expert panel believed that this level of specification is not warranted for the QIs. Patient education and counseling, although an important indicator of quality in cirrhosis, can only be estimated using patient surveys. In addition, the findings were obtained from a predominantly male veteran population in three VA hospitals and therefore the findings may not be generalizable to women, veterans seeking care in other VA hospitals, and nonveterans in other health-care systems. However, the quality scores from our study are similar to those (or are perhaps even better than) from nonveteran populations and thus provide convergent validity to our findings (14–16). Some of the low adherence rates (particularly for patients seen during the earlier years of the study) likely reflect emerging evidence over time. Nonetheless, even in the subgroup of patients who were seen in 2006–2007, we © 2014 by the American College of Gastroenterology
found wide variation in the QI rates (28.6–86%, data not shown). Our data also do not reflect the current quality of varices care. In addition to the publication and dissemination of the clinical practice guidelines, there may have been changes related to the practice settings and the commitment of facilities’ leadership and interest in cirrhosis care delivery. Although the effect of these changes on the quality of cirrhosis care in general and on varicerelated care in particular is unclear, it is possible that the quality scores may have improved since 2007. In summary, our data show that the quality of primary prophylaxis and endoscopic treatment of variceal bleeding met the accepted standards. However, significant shortfalls remain in five of eight QIs of varices care; patients received the recommended care less than two-thirds of the time even after accounting for possible justified exceptions. Our data suggest that a GI/liver specialist’s involvement may improve the quality of varices care delivered to patients with cirrhosis. CONFLICT OF INTEREST
Guarantor of the article: Fasiha Kanwal, MD, MSHS. Specific author contributions: Study concept and design, analysis and interpretation of data, and preparation of the manuscript: Paula Buchanan; study concept and design, interpretation of data, critical revision of the manuscript for important intellectual content: Jennifer R. Kramer; study concept and design, interpretation of data, critical revision of the manuscript for important intellectual content: Hashem El-Serag, Bruce Bacon, Steven Asch; acquisition of data: Youssef Assioun; study concept and design, interpretation of data, critical revision of the manuscript for important intellectual content, obtained funding, and study supervision: Fasiha Kanwal. Financial support: This material is based on work supported by the 2008 American Society of Gastrointestinal Endoscopy Quality of Care Award, the 2009 American College of Gastroenterology Clinical Research Award and in part by the Houston VA HSR&D Center of Excellence (HFP90-020). Potential competing interests: The opinions and assertions contained herein are the sole views of the authors and are not to be construed as official or as reflecting the views of the Department of Veterans Affairs. The authors declare no conflict of interest.
REFERENCES 1. Franchis R, Primignani M. Natural history of portal hypertension in patients with cirrhosis. Clin Liver Dis 2001;5:645–63. 2. Min YW, Bae SY, Gwak GY et al. A clinical predictor of varices and portal hypertensive gastropathy in patients with chronic liver disease. Clin Mol Hepatol 2012;18:178–84. 3. The North Italian Endoscopic Club for the Study and Treatment of Esophageal Varices. Prediction of the first variceal hemorrhage in patients with cirrhosis of the liver and esophageal varices. A prospective multicenter study. N Engl J Med 1988;319:983–9. 4. Carbonell N, Pauwels A, Serfaty L et al. Improved survival after variceal bleeding in patients with cirrhosis over the past two decades. Hepatology 2004;40:652–9. 5. Graham DY, Smith JL. The course of patients after variceal hemorrhage. Gastroenterology 1981;80:800–9. 6. Bernard B, Lebrec D, Mathurin P et al. Beta-adrenergic antagonists in the prevention of gastrointestinal rebleeding in patients with cirrhosis: a metaanalysis. Hepatology 1997;25:63–70.
The American Journal of GASTROENTEROLOGY
939
940
THE RED SECTION
7. Pascal JP, Cales P. Propranolol in the prevention of first upper gastrointestinal tract hemorrhage in patients with cirrhosis of the liver and esophageal varices. N Engl J Med 1987;317:856–61. 8. Sanyal AJ. Hepatic venous pressure gradient: to measure or not to measure, that is the question. Hepatology 2000;32:1175–6. 9. Khuroo MS, Khuroo NS, Farahat KL et al. Meta-analysis: endoscopic variceal ligation for primary prophylaxis of oesophageal variceal bleeding. Aliment Pharmacol Ther 2005;21:347–61. 10. Laine L, Cook D. Endoscopic ligation compared with sclerotherapy forthe treatment of esophageal variceal bleeding: a meta-analysis. Ann Intern Med 1995;123:280–7. 11. Bañares R, Albillos A, Rincón D et al. Endoscopic treatment versus endoscopic plus pharmacologic treatment for acute variceal bleeding: a metaanalysis. Hepatology 2002;35:609–15. 12. Corley DA, Cello JP, Adkisson W et al. Octreotide for acute esophageal variceal bleeding: a meta-analysis. Gastroenterology 2001;120:946–54. 13. Soares-Weiser K, Brezis M, Tur-Kaspa R et al. Antibiotic prophylaxis for cirrhotic patients with gastrointestinal bleeding. Cochrane Database Syst Rev 2002, CD002907. 14. Wilbur K, Sidhu K. Beta blocker prophylaxis for patients with variceal hemorrhage. J Clin Gastroenterol 2005;39:435–40. 15. Sorbi D, Gostout CJ, Peura D et al. An assessment of the management of acute bleeding varices: a multicenter prospective member-based study. Am J Gastroenterol 2003;98:2424–34.
16. Singh H, Targownik LE, Ward G et al. An assessment of endoscopic and concomitant management of acute variceal bleeding at a tertiary care centre. Can J Gastroenterol 2007;21:85–90. 17. Kanwal F, Kramer J, Asch SM et al. An explicit quality indicator set for measurement of quality of care in patients with cirrhosis. Clin Gastroenterol Hepatol 2010;8:709–17. 18. Kanwal F, Kramer JR, Buchanan P et al. The quality of care provided to patients with cirrhosis and ascites in the Department of Veterans Affairs. Gastroenterology 2012;143:70–7. 19. Atkins D, Best D, Briss PA et al. GRADE Working Group. Grading quality of evidence and strength of recommendations. Br Med J 2004;328:1490. 20. Wai CT, Greenson JK, Fontana RJ et al. A simple noninvasive index can predict both significant fibrosis and cirrhosis in patients with chronic hepatitis C. Hepatology 2003;38:518–26. 21. Kamath PS, Wiesner RH, Malinchoc M et al. A model to predict survival in patients with end-stage liver disease. Hepatology 2001;33:464–70. 22. Deyo RA, Cherkin DC, Ciol MA. Adapting a clinical comorbidity index for use with ICD9-CM administrative databases. J Clin Epidemiol 1992;45:613–9. 23. SAS statistical program, version 9.1 (SAS, Cary, NC). 24. Ayanian JZ, Landrum MB, Guadagnoli E et al. Specialty of ambulatory care physicians and mortality among elderly patients after myocardial infarction. N Engl J Med 2002;347:1678–86. 25. Kanwal F, Schnitzler MS, Bacon BR et al. Quality of care in patients with chronic hepatitis C virus infection: a cohort study. Ann Intern Med 2010;153:231–9.
APPENDIX Table A1. Chart validation results Variable
PPV
NPV
Kappa
Database definition
Variceal bleeding
0.70 (0.47–0.86)
0.98 (0.94–0.99)
0.72 (0.55–0.88)
At least one inpatient ICD-9 code
Varices
0.89 (0.74–0.96)
0.90 (0.84–0.94)
0.69 (0.56–0.81)
At least one inpatient ICD-9 code
Upper endoscopy
0.90 (0.81–0.95)
0.96 (0.90–0.98)
0.86 (0.79–0.94)
At least one code
NPV, negative predictive value; PPV, positive predictive value. Variceal bleeding: ICD-9: 456.0, 456.20, 456.2. Varices: ICD-9: 456.0, 456.1, 456.2, 456.20, 456.21. Upper endoscopy: ICD-9: 45.13, 45.16, 44.43, CPT: 43200, 43202, 43204, 43205, 43226, 43227, 43234, 43235, 43236, 43239, 43241, 43242, 43243, 43244, 43245, 43246, 43247, 43248, 43249, 43251, 43255, 43258, 0008T, 91110.
The American Journal of GASTROENTEROLOGY
VOLUME 109 | JULY 2014 www.amjgastro.com
THE RED SECTION
nature publishing group
The Quality of Care Provided to Patients With Varices in the Department of Veterans Affairs Paula M. Buchanan, PhD1, Jennifer R. Kramer, PhD2, Hashem B. El-Serag, MD, MPH2,3, Steven M. Asch, MD, MPH4, Youssef Assioun, MD5, Bruce R. Bacon, MD5 and Fasiha Kanwal, MD, MSHS2,3 OBJECTIVES:
Practice guidelines define the criteria and standards of care in patients with cirrhosis and varices. However, the extent to which the patients receive recommended care is largely unknown. We evaluated the quality of varices related care and factors associated with receipt of such care.
METHODS:
We conducted a retrospective cohort study of 550 patients with cirrhosis who sought care at three VA facilities between 2000 and 2007. Using administrative and clinical data, we assessed quality of varices care as measured by eight explicit Delphi panel-derived quality indicators. We also conducted a structured implicit review of patients’ medical records to explore the role of patients’ refusal, receipt of care outside the VA, or justifiable exclusions to certain care processes as explanations for non-adherence to the quality indicators.
RESULTS:
Quality scores (max. 100%) varied across individual indicators, ranging from 24.3% for upper endoscopy for varices screening to 72.4% for secondary prophylaxis for variceal bleeding. Justifiable exclusions to indicated care documented in charts were common for primary prophylaxis in patients with varices; receipt of endoscopy; and endoscopic treatment in patients with active bleeding. In contrast, significant shortfalls remained in the receipt of screening endoscopy, use of beta-blockers (in the absence of varices), and use of somatostatin analogs, antibiotics, and secondary prophylaxis in patients with variceal bleeding. Younger patients ( < 60 vs. > 60 year, odds ratio (OR) = 1.29, 95% confidence interval (CI) 1.01–1.68), those who saw a gastroenterologist (OR = 1.55, 95% CI = 1.09– 2.21), or those who were seen in the facility with academic affiliation (OR = 1.26, 95% CI = 1.01– 1.58) received higher quality care.
CONCLUSIONS:
Health-care quality, measured according to whether patients received recommended varices-related care, was suboptimal in this health-care setting. Care that included gastroenterologists was associated with high quality.
Am J Gastroenterol 2014; 109:934–940; doi:10.1038/ajg.2013.487
Gastroesophageal varices are present in almost half of the patients presenting with cirrhosis at the time of initial diagnosis (1,2). Variceal bleeding occurs in 25–35% of patients with cirrhosis (3). Up to 15% of initial bleeding episodes are fatal (4), and 70% of survivors may have recurrent bleeding (5). Despite significant developments in the management of varices, mortality from bleeding varices remains high (4). Effective and timely care can improve outcomes in patients with varices and variceal bleeding. For example, clinical studies demonstrate that both beta-blockers and endoscopic variceal ligation (EVL) are effective in preventing first episode
of variceal bleeding (6–9). Results of controlled trials show that EVL is superior to sclerotherapy for control of active bleeding, and that this superiority is consistent across all key outcomes (10). Similarly, data show that use of somotostatin analogs reduces the risk of initial treatment failure and blood transfusions in patients with active variceal bleeding (11,12). Randomized controlled studies also demonstrate the benefit of using antibiotic prophylaxis in patients with variceal bleeding (13). Despite the available evidence, existing data indicate the presence of systematic deficiencies in the care that is provided to patients
1
Center for Outcomes Research, Saint Louis University, Saint Louis, Missouri, USA; 2Michael E. DeBakey Veterans Affairs Medical Center, Houston, Texas, USA; 3Division of Gastroenterology, Department of Medicine, Baylor College of Medicine, Houston, Texas, USA; 4Department of Medicine and Health Services Research, Palo Alto VA Healthcare System, Palo Alto, California, USA; 5Department of Medicine, Saint Louis University, Saint Louis, Missouri, USA. Correspondence: Fasiha Kanwal, MD, MSHS, Michael E. DeBakey Veterans Affairs Medical Center, 2002 Holcombe Boulevard (152), VA, BCM, Houston, Texas 77030, USA. E-mail: [email protected] or [email protected] The American Journal of GASTROENTEROLOGY
VOLUME 109 | JULY 2014 www.amjgastro.com
THE RED SECTION
with varices. For example, Wilbur and Sidhu (14) found that 94% of patients with variceal bleeding had not received any primary prophylaxis. Recent data also suggest wide variations in the treatment of active bleeding across treating providers. For example, in a multicenter study from the United States including 700 patients, endoscopic intervention included banding in 41%, sclerotherapy in 36%, and both modalities in 6% of patients (15). Similarly, Singh et al. (16) found that follow-up endoscopy for secondary prophylaxis was arranged for only 65% of the patients after endoscopic treatment of the initial bleeding episode. These examples highlight that the quality of care in the management of varices may be suboptimal. However, these studies have focused on a single process each. No study has examined quality of care across the sequential processes of care—from diagnosis to treatment and preventive care—in the management of patients with varices. We recently developed and published an explicit set of evidence-based quality indicators (QIs) in cirrhosis using the modified Delphi panel approach (17). In a previous report, we used the ascite-specific QIs to assess the quality of ascitesrelated care (18). We now extend our evaluation to assess the quality of varices care in a large sample of patients with cirrhosis who sought care at three Veterans Affairs (VA) medical centers between 2000 and 2007. We also examined the relationship between patient factors and quality of care, and explored the role of justifiable exclusions, patients’ refusal, and receipt of care outside the VA health-care system as possible explanations for receipt of lower quality care.
Methods
Using an administrative and clinical database, we determined the adherence to an explicit set of varices-specific QIs in a cohort of patients with cirrhosis. We then conducted a structured implicit review of patients’ medical charts to explain findings from the explicit analysis. Quality indicators. Quality of variceal bleeding care was measured using eight specific QIs. These QIs were derived from a larger set of QIs measuring care provided to patients with cirrhosis, as described before (17). The original set included 18 QIs in the variceal bleeding care domain. Although all were rated as appropriate indicators of quality, the amount and quality of evidence as well as the size and precision of the treatment effect to support individual QIs varied. For this analysis, we included only those varices QIs that were supported by evidence from randomized trials or meta-analyses (Levels A and B per GRADE method) (19). The one exception was the “receipt of upper endoscopy” to screen for varices because it has widespread expert support from the guidelines. Table 1 lists the QIs that we examined in this study. They included three QIs measuring pre-bleeding care (receipt of upper endoscopy to screen for varices, receipt of primary prophylaxis with beta-blockers or EVL, and not receiving beta-blockers in the absence of documented varices); four QIs measuring variceal bleeding care (receipt of somatostatin analogs, antibiotics, upper © 2014 by the American College of Gastroenterology
endoscopy, and either EVL or sclerotherapy); and one QI measuring post-bleeding care (secondary prophylaxis with beta-blockers and EVL or sclerotherapy). Table 1 also shows the numerator, denominator, and time frame for each measure during which an indicated process was required to occur. For example, patients were classified as having met the “received an upper endoscopy to screen for varices” QI if they received an upper endoscopy within 12 months of cirrhosis diagnosis. Data source and study sample. We used a database derived from the VA electronic medical record system that included information for patients who sought care at 3 VA Medical Centers and their 15 affiliated clinics in VISN-15 of the Midwest United States. The database includes separate files for patient demographics, outpatient clinic and inpatient utilization, vital signs, pharmacy utilization, and laboratory data. We defined cirrhosis patients as adults with at least two ICD-9 codes for cirrhosis (571.2, 571.5, and 571.6) or at least one ICD-9 code for cirrhosis in combination with either a code for complications of cirrhosis (456.0, 456.1, 456.20, 456.21, 572.2, 572.3, 572.4, 572.8, and 789.5) or aspartate aminotransferase: platelet ratio > 2 between January 2000 and December 2007 (20). Patients had to be eligible for at least one QI to be included in the analysis. Chart validation of database-derived variables. For each QI, we developed specifications for measurement based on a combination of sources (ICD-9 codes and CPT codes, etc.) in the database. We estimated the validity of database-derived definitions of the key variables (such as varices, variceal bleeding, and upper endoscopy) by comparing them with diagnoses derived from a structured and detailed review of the patients’ medical records (Appendix Table A1). We selected 300 patients from the database for the chart validation. A trained clinician abstracted the medical record using a standardized, detailed abstraction form to determine the presence or absence of the key conditions listed above. The abstractor was blinded to the database coding of these conditions. A 10% random sample was additionally reviewed by a senior clinician. The study team discussed and resolved any discrepancies by consensus.
Statistical methods
Chart review validation. We calculated the positive predictive value (PPV), negative predictive value (NPV), and kappa coefficient for database-derived definitions in correctly identifying varices, variceal bleeding, and upper endoscopy in the medical records. PPV is the probability that the condition recorded in administrative data is actually present in the medical record, our gold standard, and NPV is the probability that the condition classified as absent based on administrative data is actually absent according to the medical record. Estimating rates of QI adherence. For each subject, we determined eligibility for the process specified in each QI (i.e., included in the denominator). We then determined the receipt of The American Journal of GASTROENTEROLOGY
935
936
THE RED SECTION
Table 1. Rates of quality indicators, defined as percentage of cirrhosis patients who received the care indicated by each indicator (numerator) among those who were eligible for the indicator (denominator) QI
Rate % (95% confidence interval) (denominator)
Quality indicators
Pre-bleeding domain 1
Among patients with cirrhosis and no documented history of variceal bleedinga, % who received an upper endoscopy within 12 months of cirrhosis diagnosis
24.3 (20.2-28.3) (441)
2
Among patients with cirrhosis and no documented history of variceal bleedingb, % who did not receive non-selective beta-blockers for primary variceal prophylaxis
72.4 (68.0–76.9) (388)
3
Among patients with cirrhosis, no documented history of variceal bleedingc, but with varices on endoscopy, % who received either non-selective beta-blockers or esophageal variceal ligation (EVL) within 1 month of varices diagnosis (including non-selective beta-blockers before varices diagnosis)
60.4 (50.2–70.7) (91)
4
Among patients with cirrhosis who have variceal bleeding on or after cirrhosis index date, % who receive somatostatin or analogs (somatostatin, octreotide, terlipressin) within the index hospitalization
59.3 (50.1–68.5) (113)
5
Among patients with cirrhosis who have variceal bleeding on or after cirrhosis index date, % who receive an upper endoscopy within 24 hours of presentation
74.8 (66.9–82.7) (119)
6
Among patients with cirrhosisd and variceal bleeding, % who receive EVL or sclerotherapy within 1 week of varcieal bleeding date
63.5 (51.3–75.7) (63)
7
Among patients with cirrhosis who have variceal bleeding, % who receive antibiotic prophylaxis during hospitalization
37.7 (28.7–46.8) (114)
Bleeding-related domain
Secondary prevention domain 8
Among patients with cirrhosis who have variceal bleeding who survived for 3 months post bleeding, % who received either nonselective beta-blockers and/or EVL or sclerotherapy within 3 months of varcieal bleeding date
72.4 (62.6–82.0) (87)
a
No bleeding 1 year before cirrhosis index date through endoscopy date. No bleeding for 1 year before cirrhosis index date and no varices from 1 year before to 1 year after cirrhosis index date. c No bleeding until 1 day after the endoscopy date. d Removed patients with gastric varices codes from the denominator. b
recommended care within the specified time frame. We calculated the rate of each QI as the percentage of subjects who received the recommended care from among those subjects who were eligible for the QI. Implicit chart review. Using the chart validation sample described above, we conducted a structured review of patients’ medical records to explore justifiable explanations for non-adherence to the indicated care processes that were not covered by the QI specifications above. These included, but were not restricted to, receipt of care outside the VA, being too sick to receive indicated care, or patients’ refusal. For each QI, we only reviewed cases that were eligible for, but did not receive the care as indicated by the QI. Predictors of quality care. We used multivariable analysis to identify factors associated with receipt of varices-related care. The dichotomous dependent variable was receiving the recommended care for each QI. Each patient had a row of data for each QI for which they were eligible. As our data were hierarchically structured— with QIs clustered within patients—we constructed multilevel generalized linear mixed models (GLIMMIX in SAS (SAS, Cary, NC)) with random effects for the patients to determine predictors The American Journal of GASTROENTEROLOGY
of varices care. The independent variables included demographic factors (age, race), liver disease severity-related characteristics (model for end-stage liver disease or MELD score (21), presence of ascites), and medical comorbidities (Deyo index) (22). Table 2 provides the full list, operational definitions, and categorization of these patient factors. We defined patients as having seen a specialist if she/he had a visit (either inpatient or outpatient, based on the QI) with a gastroenterologist during the observation period (i.e., time frame) for any of the indicated QIs. We assigned each patient a facility where the patient was seen most frequently. We created a variable indicating the academic affiliation of the facilities. We reported our results as odds ratios (ORs) with 95% confidence intervals (CIs), where ORs are the odds of receiving a recommended QI for which a patient was eligible. We considered P < 0.05 as statistically significant. We used SAS (version 9.2) for all statistical analyses (SAS) (23).
Results
Demographics and clinical characteristics. We identified 550 cirrhosis patients who were eligible for at least one variceal bleeding VOLUME 109 | JULY 2014 www.amjgastro.com
THE RED SECTION
Table 2. Characteristics of 550 cirrhosis patients who were eligible for at least one variceal bleeding care quality indicator Variable
Distribution
Demographics Age in years (mean, s.d.)
54.1 (9.4)
Race (N, %) Caucasian
365 (66.4)
African American
87 (15.8)
Unknown
98 (17.8)
Gender (N, %) Male
544 (98.1)
Female
6 (1.1)
Clinical characteristics Etiology of cirrhosis (N, %) Hepatitis C
254 (46.2)
Alcohol
181 (32.9)
Hepatitis B
Results of chart validation. The PPV, NPV, and kappa for our database definition of varices were 89% (95% CI = 74–96%), 90% (84–94%), and 69% (56–81%), respectively. Similarly, the definition of an upper endoscopy had a PPV, NPV, and kappa of 90% (95% CI = 81–95%), 96% (95% CI = 90–98%), and 86% (95% CI = 79–94%), respectively. The PPV for variceal bleeding was 70% (95% CI = 47–86%), whereas the NPV and kappa were 97% (94–99%) and 71% (95% CI = 55–88%), respectively (Appendix Table A1). Rates of QIs. The rate of meeting QIs varied (Table 1). For example, whereas 59.3% (95% CI = 50.1–68.5%) of patients received somatostatin analogs at the time of variceal bleeding, only onethird (37.7%, 95% CI = 28.7–46.8%) received antibiotic prophylaxis during that episode. Similarly, 24.3% (95% CI = 20.2–28.3%) of patients had an upper endoscopy during the first year after their cirrhosis diagnosis. However, 60.4% (95% CI = 50.2–70.7%) of the patients with varices received either beta-blockers or EVL for primary prophylaxis of variceal bleeding. Overall, fewer than 16% (95% CI = 13–19%) patients received all indicated care.
2 (0.4)
Others
113 (20.6)
Complications of cirrhosis (N, %) Ascites
210 (38.2)
Variceal bleeding
123 (22.4)
Hepatic encephalopathy
117 (21.3)
Baseline MELD score (mean, s.d.)a
12.6 (5.3)
Comorbidity index (mean, s.d.)
2.2 (0.9)
Health-care utilization Outpatient visit days per yearb (median, IQR) b
21.4 (12.4–33.2)
At least one hospitalization (N, %)
369 (67.1)
Seen by a gastroenterologist (N, %)
420 (76.4)
IQR=interquartile range. a Data to allow MELD score calculation were available for 502 (91%) of our sample. Baseline MELD is the score closest to the cirrhosis index date. b Outpatient visits or hospitalization occurring after cirrhosis index date.
QI (Table 2). The mean age of our study subjects was 54 (s.d. = 9.4) years, almost all were male, and majority were white (66.4%). Cirrhosis was related to chronic viral hepatitis C and alcohol in 46.2 and 32.9% of patients, respectively. Over a third (38.2%) of the patients had ascites, 22.4% had variceal bleeding, and 21.3% had hepatic encephalopathy. The mean MELD score and Deyo index were 12.6 (s.d. = 5.3) and 2.2 (s.d. = 0.9), respectively. Patients were followed for a mean of 3.3 (s.d. = 1.8) years. The majority of the patients (67.1%) had at least one hospitalization after their cirrhosis diagnosis with a median of 21.4 (IQR = 12.4–33.3) outpatient visits per year. Approximately 76% of patients were seen by gastroenterologists during the time frame specified by at least one of the QIs. © 2014 by the American College of Gastroenterology
Reasons for non-adherence to QIs based on chart review. As shown in Table 3, the proportion of patients with possible justifiable exceptions (i.e., reasons for not receiving care) varied across the QIs. Possible exceptions to indicated care documented in charts were particularly common for “receipt of primary prophylaxis”, “receipt of upper endoscopy in patients with variceal bleeding”, and “receipt of EVL or sclerotherapy for variceal bleeding” QIs. For “receipt of primary prophylaxis” QI, only four patients in our chart review subsample were eligible for but did not meet the QI. Of these, two had small varices, one had low blood pressure (contraindicating beta-blockers but not EVL), and one had metastatic cancer. Similarly, for “receipt of upper endoscopy in patients with variceal bleeding”, three of the four patients received upper endoscopy before the start of eligibility time frame (before admission) and one died within a day of presenting with variceal bleeding. Expanding the time frame to include endoscopy procedures that occur within a few days before the eligibility date (admission date in this case) would capture these patients. Six of the eight chart review patients who did not receive EVL or sclerotherapy for variceal bleeding had justifiable exceptions; three bled from either gastric varices or portal hypertensive gastropathy, one did not have any bleeding, one was too sick, and one received recommended care at an outside facility. In contrast, the majority of patients included in our implicit review for other QIs, including “receipt of upper endoscopy to screen for varices”, “not receiving beta-blockers in the absence of documented varices”, “receipt of somatostatin analogs and antibiotics for variceal bleeding”, and “secondary prophylaxis with betablockers and EVL or sclerotherapy”, did not have an explanation for not receiving the recommended care (classified as having “no documented reason” in Table 3). Predictors of quality care. Results of the multilevel regression analysis are presented in Table 4. Younger patients, those who saw The American Journal of GASTROENTEROLOGY
937
938
THE RED SECTION
Table 3. Reasons for not receiving recommended varices care—results of the implicit chart review in 61 patients with cirrhosis Quality indicators (N of cases included in the implicit review for the indicator)
Number (percentage) of patients No documented reason
Patient refusal
Outside care
Too sick
Other reasons
Endoscopy to screen for varices (n=42)
30 (71.4)
2 (4.7)
0 (0)
1 (0.2)
9 (21.4)
Not receiving beta-blockers in the absence of documented varices (n=15)
14 (93.3)
0 (0)
0 (0)
0 (0)
1 (0.07)
Primary prophylaxis for documented varices (n=4)
0 (0)
0 (0)
0 (0)
1 (25)
3 (75.0)
5 (71.4)
0 (0)
0 (0)
0 (0)
2 (28.5)
0 (0)
0 (0)
0 (0)
1 (25.0)
3 (75.0)
Variceal ligation or sclerotherapy in patients with variceal bleeding (n=8)
2 (25.0)
0 (0)
0 (0)
1 (12.5)
5 (62.5)
Antibiotics in patients with variceal bleeding (n=12)
9 (75.0)
0 (0)
0 (0)
1 (0.83)
2 (16.6)
Secondary prophylaxis (n=4)
3 (75.0)
0 (0)
1 (25.0)
0 (0)
0 (0)
Somatostatin analogs in patients with variceal bleeding (n=7) Endoscopy in patients with variceal bleeding (n=4)
The number of patients who were eligible for but did not meet the individual quality indicators (QIs) are listed with each QI.
a gastroenterologist, or those who were seen in the facility with academic affiliation were more likely to meet the QIs for which they are eligible. For example, patients who were younger than 60 years were 29% more likely to meet the QIs for which they are eligible compared with older patients (OR = 1.29, 95% CI = 1.01– 1.68). If the patient saw a gastroenterologist, then their odds of receiving the recommended care were 55% higher (OR = 1.55, 95% CI = 1.09–2.21), whereas those treated at an academic facility had 26% higher odds of receiving recommended care (OR = 1.26, 95% CI = 1.01–1.58). Removing the QIs that were sensitive to the database limitations (with > 50% rate of justifiable exceptions) did not change the direction or magnitude of the results above.
Discussion
Our study has three key findings. First, we found that the quality of varices care, measured according to whether patients received recommended services, varied substantially across the process of care examined. We found that most patients with documented varices received recommended primary prophylaxis with either nonselective beta-blockers or EVL. Similarly, 75% of patients who presented with variceal bleeding received an endoscopy within 24 h of presentation and most (64%) underwent recommended EVL or sclerotherapy for management of variceal bleeding. In contrast, less than a third of the patients underwent an endoscopy to screen for varices within a year of cirrhosis diagnosis and fewer than 40% received antibiotics during admission for variceal bleeding. Care for other processes fell in between the two extremes. Somatostatin analogs were used in about 60% of patients who presented with acute bleeding, and a quarter of patients did not receive recommended secondary prophylaxis against re-bleeding. These deficits persisted even after accounting for possible justified exceptions in our chart review. Second, we found that automated data underestimated the rates for the measures for which justifiable exceptions for failing to The American Journal of GASTROENTEROLOGY
Table 4. Results of multivariable regression analysis showing the association of demographic and clinical factors with receipt of variceal bleeding care in patients with cirrhosis Variable
Odds ratio
95% Confidence interval
1.29
1.01–1.68
Demographics Age in years (reference ≥60) Race (reference white)
—
—
African American
0.74
0.54–1.01
Unknown race
0.87
0.65–1.15
MELD score (reference ≥10)
1.06
0.84–1.34
Presence of ascites (reference no)
1.04
0.83–1.31
—
—
0/1
1.14
0.86–1.50
2
1.08
0.74–1.21
Seen by a gastroenterologist (reference no.)
1.55
1.09–2.21
Seen in academic facility (reference no.)
1.26
1.01–1.58
Clinical characteristics
Deyo index (reference ≥3)
Health-care system factors
The odds ratio indicate the odds of receiving recommended care for which the patient was eligible. Continuous variables are categorized on the basis of their distribution in the data (i.e., percentiles).
provide the recommended care were common. We found that these exceptions occurred frequently and had a large effect on measured performance for three QIs: receipt of primary prophylaxis, receipt of upper endoscopy in patients with variceal bleeding, and receipt of EVL or sclerotherapy for variceal bleeding QIs. We therefore VOLUME 109 | JULY 2014 www.amjgastro.com
THE RED SECTION
would not recommend the application of these measures using automated data unless a chart review or other validated measures were also used to improve accuracy. However, manual chart reviews may be too inefficient and expensive to be disseminated widely, particularly when used for tracking changes in quality care over time. Other alternatives would be the natural language processing of free text notes or creating new standard codes that explain exceptions or patient preferences for treatment decisions as endorsed by the Centers for Medicare and Medicaid Services’ Physician Quality Reporting System. These codes could be associated with clinical encounters or placed in problem lists to allow capture by automated searches. Third, we found that being seen by gastroenterologists was an important determinant of receipt of quality care in patients with (or those at risk for) varices. These results are consistent with the available data that care by specialists within their area of expertise may provide the best chance at improving the process and/or outcomes of care (24–25). As endoscopy is performed mostly by gastroenterologists, we considered the possibility that the effect of seeing a gastroenterologist on the overall quality of varices care may be mediated solely through its influence on measures related to endoscopic care. However, we found a consistently positive association between being seen by gastroenterologists and each measure separately, although some of these associations were stronger than the others (data not shown). Our data suggest that specialist involvement may improve the quality of care delivered to patients with cirrhosis and varices, although future research will be needed to determine whether this effect is only mediated through direct patient care or could be achieved through tele-consultation with primary care providers. Our study has several limitations. We cannot make causal inferences between QIs and the predictors, given the observational retrospective nature of the study design. Similarly, as the database lacked relevant information, we could not identify specific characteristics of specialists or of the academic setting that explained the respective associations with better care. We also could not assess certain aspects of care such as beta-blocker dose titration and duration, number of bands placed during EVL, patient education, or counseling regarding alcohol use. As some of these parameters (such as dose titration and number of bands placed) are tailored to a given patient’s clinical condition, the expert panel believed that this level of specification is not warranted for the QIs. Patient education and counseling, although an important indicator of quality in cirrhosis, can only be estimated using patient surveys. In addition, the findings were obtained from a predominantly male veteran population in three VA hospitals and therefore the findings may not be generalizable to women, veterans seeking care in other VA hospitals, and nonveterans in other health-care systems. However, the quality scores from our study are similar to those (or are perhaps even better than) from nonveteran populations and thus provide convergent validity to our findings (14–16). Some of the low adherence rates (particularly for patients seen during the earlier years of the study) likely reflect emerging evidence over time. Nonetheless, even in the subgroup of patients who were seen in 2006–2007, we © 2014 by the American College of Gastroenterology
found wide variation in the QI rates (28.6–86%, data not shown). Our data also do not reflect the current quality of varices care. In addition to the publication and dissemination of the clinical practice guidelines, there may have been changes related to the practice settings and the commitment of facilities’ leadership and interest in cirrhosis care delivery. Although the effect of these changes on the quality of cirrhosis care in general and on varicerelated care in particular is unclear, it is possible that the quality scores may have improved since 2007. In summary, our data show that the quality of primary prophylaxis and endoscopic treatment of variceal bleeding met the accepted standards. However, significant shortfalls remain in five of eight QIs of varices care; patients received the recommended care less than two-thirds of the time even after accounting for possible justified exceptions. Our data suggest that a GI/liver specialist’s involvement may improve the quality of varices care delivered to patients with cirrhosis. CONFLICT OF INTEREST
Guarantor of the article: Fasiha Kanwal, MD, MSHS. Specific author contributions: Study concept and design, analysis and interpretation of data, and preparation of the manuscript: Paula Buchanan; study concept and design, interpretation of data, critical revision of the manuscript for important intellectual content: Jennifer R. Kramer; study concept and design, interpretation of data, critical revision of the manuscript for important intellectual content: Hashem El-Serag, Bruce Bacon, Steven Asch; acquisition of data: Youssef Assioun; study concept and design, interpretation of data, critical revision of the manuscript for important intellectual content, obtained funding, and study supervision: Fasiha Kanwal. Financial support: This material is based on work supported by the 2008 American Society of Gastrointestinal Endoscopy Quality of Care Award, the 2009 American College of Gastroenterology Clinical Research Award and in part by the Houston VA HSR&D Center of Excellence (HFP90-020). Potential competing interests: The opinions and assertions contained herein are the sole views of the authors and are not to be construed as official or as reflecting the views of the Department of Veterans Affairs. The authors declare no conflict of interest.
REFERENCES 1. Franchis R, Primignani M. Natural history of portal hypertension in patients with cirrhosis. Clin Liver Dis 2001;5:645–63. 2. Min YW, Bae SY, Gwak GY et al. A clinical predictor of varices and portal hypertensive gastropathy in patients with chronic liver disease. Clin Mol Hepatol 2012;18:178–84. 3. The North Italian Endoscopic Club for the Study and Treatment of Esophageal Varices. Prediction of the first variceal hemorrhage in patients with cirrhosis of the liver and esophageal varices. A prospective multicenter study. N Engl J Med 1988;319:983–9. 4. Carbonell N, Pauwels A, Serfaty L et al. Improved survival after variceal bleeding in patients with cirrhosis over the past two decades. Hepatology 2004;40:652–9. 5. Graham DY, Smith JL. The course of patients after variceal hemorrhage. Gastroenterology 1981;80:800–9. 6. Bernard B, Lebrec D, Mathurin P et al. Beta-adrenergic antagonists in the prevention of gastrointestinal rebleeding in patients with cirrhosis: a metaanalysis. Hepatology 1997;25:63–70.
The American Journal of GASTROENTEROLOGY
939
940
THE RED SECTION
7. Pascal JP, Cales P. Propranolol in the prevention of first upper gastrointestinal tract hemorrhage in patients with cirrhosis of the liver and esophageal varices. N Engl J Med 1987;317:856–61. 8. Sanyal AJ. Hepatic venous pressure gradient: to measure or not to measure, that is the question. Hepatology 2000;32:1175–6. 9. Khuroo MS, Khuroo NS, Farahat KL et al. Meta-analysis: endoscopic variceal ligation for primary prophylaxis of oesophageal variceal bleeding. Aliment Pharmacol Ther 2005;21:347–61. 10. Laine L, Cook D. Endoscopic ligation compared with sclerotherapy forthe treatment of esophageal variceal bleeding: a meta-analysis. Ann Intern Med 1995;123:280–7. 11. Bañares R, Albillos A, Rincón D et al. Endoscopic treatment versus endoscopic plus pharmacologic treatment for acute variceal bleeding: a metaanalysis. Hepatology 2002;35:609–15. 12. Corley DA, Cello JP, Adkisson W et al. Octreotide for acute esophageal variceal bleeding: a meta-analysis. Gastroenterology 2001;120:946–54. 13. Soares-Weiser K, Brezis M, Tur-Kaspa R et al. Antibiotic prophylaxis for cirrhotic patients with gastrointestinal bleeding. Cochrane Database Syst Rev 2002, CD002907. 14. Wilbur K, Sidhu K. Beta blocker prophylaxis for patients with variceal hemorrhage. J Clin Gastroenterol 2005;39:435–40. 15. Sorbi D, Gostout CJ, Peura D et al. An assessment of the management of acute bleeding varices: a multicenter prospective member-based study. Am J Gastroenterol 2003;98:2424–34.
16. Singh H, Targownik LE, Ward G et al. An assessment of endoscopic and concomitant management of acute variceal bleeding at a tertiary care centre. Can J Gastroenterol 2007;21:85–90. 17. Kanwal F, Kramer J, Asch SM et al. An explicit quality indicator set for measurement of quality of care in patients with cirrhosis. Clin Gastroenterol Hepatol 2010;8:709–17. 18. Kanwal F, Kramer JR, Buchanan P et al. The quality of care provided to patients with cirrhosis and ascites in the Department of Veterans Affairs. Gastroenterology 2012;143:70–7. 19. Atkins D, Best D, Briss PA et al. GRADE Working Group. Grading quality of evidence and strength of recommendations. Br Med J 2004;328:1490. 20. Wai CT, Greenson JK, Fontana RJ et al. A simple noninvasive index can predict both significant fibrosis and cirrhosis in patients with chronic hepatitis C. Hepatology 2003;38:518–26. 21. Kamath PS, Wiesner RH, Malinchoc M et al. A model to predict survival in patients with end-stage liver disease. Hepatology 2001;33:464–70. 22. Deyo RA, Cherkin DC, Ciol MA. Adapting a clinical comorbidity index for use with ICD9-CM administrative databases. J Clin Epidemiol 1992;45:613–9. 23. SAS statistical program, version 9.1 (SAS, Cary, NC). 24. Ayanian JZ, Landrum MB, Guadagnoli E et al. Specialty of ambulatory care physicians and mortality among elderly patients after myocardial infarction. N Engl J Med 2002;347:1678–86. 25. Kanwal F, Schnitzler MS, Bacon BR et al. Quality of care in patients with chronic hepatitis C virus infection: a cohort study. Ann Intern Med 2010;153:231–9.
APPENDIX Table A1. Chart validation results Variable
PPV
NPV
Kappa
Database definition
Variceal bleeding
0.70 (0.47–0.86)
0.98 (0.94–0.99)
0.72 (0.55–0.88)
At least one inpatient ICD-9 code
Varices
0.89 (0.74–0.96)
0.90 (0.84–0.94)
0.69 (0.56–0.81)
At least one inpatient ICD-9 code
Upper endoscopy
0.90 (0.81–0.95)
0.96 (0.90–0.98)
0.86 (0.79–0.94)
At least one code
NPV, negative predictive value; PPV, positive predictive value. Variceal bleeding: ICD-9: 456.0, 456.20, 456.2. Varices: ICD-9: 456.0, 456.1, 456.2, 456.20, 456.21. Upper endoscopy: ICD-9: 45.13, 45.16, 44.43, CPT: 43200, 43202, 43204, 43205, 43226, 43227, 43234, 43235, 43236, 43239, 43241, 43242, 43243, 43244, 43245, 43246, 43247, 43248, 43249, 43251, 43255, 43258, 0008T, 91110.
The American Journal of GASTROENTEROLOGY
VOLUME 109 | JULY 2014 www.amjgastro.com
