Eur J Clin Pharmacol (2016) 72:631–639 DOI 10.1007/s00228-016-2022-4
PHARMACOEPIDEMIOLOGY AND PRESCRIPTION
Hospital re-admission associated with adverse drug reactions in patients over the age of 65 years Laurent Hauviller 1 & Frédéric Eyvrard 1 & Valérie Garnault 2 & Vanessa Rousseau 1 & L. Molinier 2 & Jean Louis Montastruc 1 & Haleh Bagheri 1
Received: 12 October 2015 / Accepted: 8 February 2016 / Published online: 17 February 2016 # Springer-Verlag Berlin Heidelberg 2016
Abstract Context Adverse drug reactions (ADRs) are responsible for 5 % of hospital admissions, but hospital re-admission induced by ADRs remains poorly documented. Objective The aim of this study was to estimate the rate of hospital re-admission and the factors associated with readmission in the patients over the age of 65 years. Secondary, we described the characteristics of cases of ADRs leading to re-admission for drugs other than chemotherapy agents. Methods Data were extracted from hospital discharge summaries provided by the Department of Medical Information of Toulouse University Hospital. All patients over the age of 65 years admitted to the hospital in 2010 for an ADR, identified from ICD-10 codes, were selected. All subsequent admissions of members of this cohort within 1 year of discharge following the index admission were reviewed retrospectively. The risk factors associated with hospital re-admission for ADRs were analyzed. Medical records were used for descriptive analysis of re-admission due to drugs other than chemotherapy agents.
Results We found that 553 of the 1000 patients admitted for ADRs in 2010 were re-admitted to hospital within 1 year. Among them, 87 cases were re-admitted for ADRs (estimated rate of 87/1000 re-admission for an ADR within 1 year). A comparison of the patients re-admitted for ADRs (n = 87) with those of patients re-admitted for other causes (n = 410) suggested that only cancer increased the risk of re-admission for ADRs (OR = 7.69 [4.59–12.88] 95 % CI). ADRs due to the same drug combination were the suspected cause of repeat admission in half the cases (other than chemotherapy). Hospital re-admission was considered avoidable in four cases (22 %). Conclusion This study shows an estimated rate of readmission for an ADR around 87/1000 within 1 year, and the same drug combination were the suspected cause of repeat admission in half the cases. At least, 11 % of cases were avoidable.
Electronic supplementary material The online version of this article (doi:10.1007/s00228-016-2022-4) contains supplementary material, which is available to authorized users.
Introduction
* Haleh Bagheri [email protected]
1
Service de Pharmacologie Médicale et Clinique, Centre Midi-Pyrénées de Pharmacovigilance, de Pharmacoépidémiologie et d’Informations sur le Médicament, Faculté de Médecine de l’Université Paul Sabatier-, INSERM U1027, 37 Allées Jules Guesde, 31000 Toulouse, France
2
Département d’Informations Médicales, CHU Toulouse, Hôtel Dieu St Jacques, 2 Rue Viguerie, 31300 Toulouse, France
Keywords Adverse drug reactions . Pharmacovigilance . Hospital re-admission . Avoidability
In western countries, several studies have investigated rates of hospital admission due to adverse drug reactions (ADRs). These studies have reported a prevalence of ADRs leading to hospitalization of 0.16 to 15.7 %, with an overall median prevalence of 5.3 % [1–3]. The large variation depends on the method of the study (retrospective or prospective) and the characteristics of included patients. In a recent review, Bouvy et al. suggested that the median rate of hospital admissions for ADRs was 3.5 % [4]. A higher rate would be expected in cancer patients (about 12 %), due to comorbidity and the toxicity of chemotherapy drugs [5, 6]. Likewise, the elderly
632
(people over the age of 65 years) are the largest consumers of healthcare and the recipients of more than half the drugs prescribed. It is therefore unsurprising that older people suffer more ADRs and are seven times more likely to require hospitalization due to an ADR. The incidence of ADRs leading to hospital admission increases to 9.8 % for those over the age of 75 years [2, 7–9]. Some of these ADRs leading to hospital admission should be avoidable. Some studies estimated that as many as 30–40 % of cases might have been avoided [9, 10]. Despite the major impact of hospital re-admission for ADRs on patient care and the burden placed on already overstretched hospitals, little is known about rates of re-admission to hospital due to ADRs. Two studies have suggested that between one fifth and one third of all re-admissions to hospital may be due to an ADR [8, 11]. Zhang suggested that the male gender, the duration of initial hospital stay, and Charlson comorbidity score (>7) were risk factors for repeat admission for ADRs, whereas Davies found that age increased the risk of re-admission for ADRS [8, 11]. In France, Bonnet-Zamponi et al. studied the impact of interventions designed to decrease hospital re-admission rates for patients aged 70 or over during a 6-month period following their discharge. They reported that 37.9 % of re-admissions were due to ADRs, but ADRs were not necessarily the cause of the initial hospitalization of these patients [12]. Taken into account, the scarcity of data, in 2010, we conducted a feasibility study in two units (a geriatric ward and an emergency ward) of Toulouse University Hospital over a period of 6 months to estimate the rate of hospital re-admission for ADRs in patients aged 65 and over. The first data suggested that 60 of the 556 patients admitted for ADRs in the first half of 2010 were re-admitted within 6 months of discharge, and an ADR was the reason for readmission in six of these patients [13]. We then undertook a larger study and extended our previous one to all units of Toulouse University Hospital, for a longer period (1 year of inclusion in 2010 and 1 year of follow-up of patients after discharge), to assess more accurately the rate of hospital re-admission and the factors associated with re-admission in the same population (patients aged >65 years). We also assessed the characteristics of the ADRs leading to repeat admissions, through a descriptive analysis. For this purpose, we choose to restrict the analysis to all readmissions associated with drugs other than the antineoplastic agents administered for cancer since it is well known that these drugs are involved in repeat hospital admissions and that an effective alternative treatment is lacking [5, 6].
Methods Design We carried out a retrospective cohort study at Toulouse University Hospital (2880 beds, in South West France).
Eur J Clin Pharmacol (2016) 72:631–639
Setting and population The data were obtained from the French hospital discharge database (Programme de Medicalisation des Systèmes d’Information, PMSI) and were provided by the Department of Medical Information of the hospital. Individual hospitalizations are entered into this database in the form of standardized hospital discharge summaries. In recent years, the PMSI and data from the national health insurance system have been linked, providing a useful tool for estimating the association between drug exposure and ADRs, e.g., valvulopathy due to benfluorex or thrombotic events due to the new generation of oral contraceptives [14, 15]. The data recorded include basic patient characteristics (gender, date of birth, and permanent patient identification number), days of admission and discharge, and the main and all related diagnoses. Diagnoses were encoded according to International Classification of Diseases, 10th revision (ICD-10), including all codes related to drug-induced ADRs, as used in our previous study [16, 17]. We used the definition of an ADR proposed by Edwards and Aronson [18]. The inclusion criteria were all admissions to Toulouse University Hospital of patients over the age of 65 years, over a period of 1 year (January 1–December 31, 2010), for which the discharge summary included a relevant ICD-10 code as the main or related diagnosis (sample A). We then used the permanent identification numbers of the patients selected in this first round of screening to identify all other re-admissions to Toulouse University Hospital for these patients within 1 year of discharge after the index admission (over a 1-year period from day 1 of release after the first hospitalization) (sample B, see Fig. 1). From this sample, we then selected the patients with relevant ICD-10 terms for the main or related diagnosis on the discharge summary for the repeat admission (sample C). Finally, we ensured that the ADR was correctly identified and was the cause of both initial admission and re-admission to hospital (rather than occurring during hospitalization), by checking the data for the patients in sample C, in the medical reports available via the hospital’s medical software (called ORBIS). We excluded cases for which the data noted in these medical records were insufficient for the assessment of ADRs corresponding to relevant ICD-10 terms and leading to hospital admission.
Data analysis In accordance with published findings, we investigated the association with ADR-related hospital re-admission of the following factors [7, 11]: patient age and gender, duration of the first hospitalization, and comorbidity according to the Charlson index as estimated from ICD-10 codes, for the main or related diagnosis of the corresponding hospital admission [19]. Given the risk of repeat admission associated with chemotherapy-induced ADRs, we analyzed the characteristics
Eur J Clin Pharmacol (2016) 72:631–639
Fig. 1 Selection of patients in a French hospital discharge database (Programme de Medicalisation des Systèmes d’Information, PMSI) using ICD-10 code related to ADRs during index admission and those re-admitted within 1 year of discharge from index admission
of patients re-admitted for chemotherapy-induced ADRs separately from those of patients re-admitted for ADRs associated with other types of drugs. Secondary, we carried out a descriptive analysis of cases of re-admission due to ADRs relating to drugs other than chemotherapy agents used for cancer treatment. For each case, we collected the following data: demographic data (age, gender), number of hospital admissions corresponding to ADRs (relevant ICD-10 code), and the suspected drug according to the anatomical-therapeutic classification [20], and hospital management (withdrawal or not of the suspected drug) and its pattern of prescription (prescription in an outpatient or hospital setting).
633
Given the retrospective nature of data collection in this study and the impossibility of knowing the time between drug intake and the occurrence of the ADR, we did not assess the causal relationship between ADRs and the suspected drug according to the French causality assessment method [21]. This method includes five scores (excluded, doubtful, possible, probable, certain) based on chronological (delay of occurrence of ADR, rechallenge) and semeiological (other explanations for ADRs) data. A preventability score was used for the suspected ADRs, according to Olivier’s scale with four categories of preventability of ADRs (Bpreventable,^ Bpotentially preventable,^ Bunclassable,^ Bnot preventable^), depending on data related to patient (risk factors), drug administration (error, interactions…), and possible alternative [22]. For the same reasons, we decided to focus only on cases for which the score obtained suggested that the ADR was Bpreventable.^ Statistical analysis was performed with SAS version 9.3 (SAS Institute, Cary, NC). A descriptive analysis was first performed on the characteristics of patients admitted or readmitted for ADRs. Age at the first hospital admission was analyzed both as a continuous variable and as a categorical variable: [65–70], [71–75], [76–80], and more than 80 years. The duration of the first hospitalization for ADR was classified into two categories: 7 days or less and more than 7 days as suggested by Davies [11]. Charlson index, estimated from the ICD-10 codes for the main or a related diagnosis for the corresponding hospital admission [19], was studied first as a binary variable: patients without comorbidity and patients with at least one comorbid condition. We then focused on the most frequent comorbid conditions in these patients: cancer, diabetes, heart failure, or kidney failure. We carried out a multivariate logistic regression analysis for categorical variables to compare the characteristics of patients re-admitted for ADRs with those re-admitted for other reasons. We began by inserting the variables with p values below 0.25 in univariate analyses. A descending, stepwise method was then used to exclude variables with p values below 0.05 from the final model. A survival analysis (Kaplan-Meier method) was performed to take into account the time between two hospital admissions in the two groups of patients: those re-admitted for ADRs and those re-admitted for other causes. This interval was defined as the time between the end of the first hospitalization, with the indication of an ICD-10 code relating to ADRs on the discharge report and the beginning of the next hospitalization. If the Bevent^ considered was re-admission for an ADR, then an absence of such readmission for an ADR within 1 year of the initial discharge was considered to be a Bnon-event.^ Logistic regression analysis was used to identify the risk factors associated with re-admission for an ADR (with an ICD-10 code relating to ADRs) after the initial hospitalization.
634
Results General data During 1 year from January 1 to December 31, 2010, data extraction from PMSI led to the identification of 1000 patients with at least one ICD-10 code relating to ADRs. The mean age of these patients was 77.6 years (range 65–104, SD = 7.4) and 511 (51.5 %) were female. In total, 1331 ICD-10 codes relating to ADRs were found in our sample of 1000 patients, mostly as related diagnoses (n = 1171, 88 %). The four most common codes were druginduced aplastic anemia (D61.1, n = 208, 15.6 %), unspecified adverse effect of drug (T88.7, n = 200, 15.0 %), adverse effects of anticoagulants (Y44.2, n = 171, 12.8 %), and agranulocytosis (D70, n = 72 5.4 %). A single ICD-10 code relating to ADR was found in the records of 812 patients, two codes were found for 122 patients and three were found for 32 others (34 patients had more than three ICD-10 codes relating to ADRs). About half the patients (n = 553) were re-admitted within 1 year of discharge following the index admission, for several causes. An ICD-10 code relating to ADRs was recorded for 143 of these patients. However, an analysis of the medical reports of these 143 patients led to exclude 56 cases for the following reasons: medical reports unavailable from the medical software for seven patients, absence of description of the ADRs corresponding to the ICD-10 code in the summary discharge letter for eight patients, or, in 41 patients, the ADR related to relevant ICD-10 code was present in the patient’s medical history but was not responsible for the most recent hospital admission (e.g., history of osteoporosis induced by corticosteroids). Analysis of factors associated with re-admission for ADRs A comparison of the characteristics (Table 1) of patients readmitted for ADRs (n = 87) with those of patients re-admitted for other causes (n = 410) found that only cancer increased the risk of re-admission for ADRs (OR = 7.69 [4.59–12.88]). No significant difference was found between the two groups for gender or duration of the first hospitalization (days). However, the mean age (years) of patients re-admitted for ADRs (74.0, SD = 6, range 65–89) was significantly lower than that of patients re-admitted for other causes (77.5, SD = 7.4, range 66–104) (Z test, p < 0.0001). The mean Charlson comorbidity score was higher in patients re-admitted for ADRs (5.6, SD = 1.9, range = 2–10) than in those re-admitted for other causes (4.6, SD = 1.9, range = 2–12) (Z test, p < 0.0004). Characteristics of the patients re-admitted for ADRs In total, 87 patients according to the inclusion criteria were readmitted for ADRs. According to the survival curve (see Supplementary Material Table 2), the probability of re-
Eur J Clin Pharmacol (2016) 72:631–639
admission for an ADR increased to 15 % within 4 months, thereafter remaining constant, at about 18 % (from 4 to 12 months). Eleven of the patients re-admitted for ADRs died during their hospital stay. Eight of these patients suffered from agranulocytosis due to chemotherapy. As described in the method, we analyzed the characteristics of patients hospitalized for ADRs associated with chemotherapy at the first admission (group 1) separately from those of patients hospitalized for ADRs associated with other types of drugs at their first admission (group 2): &
&
Group 1—Patients hospitalized for ADRs associated with chemotherapy at the first hospital admission: there were 69 patients in this category, with a mean age of 73.1 years (range = 65–89, SD = 5.7), a sex ratio of 1.5, and a mean number of re-admissions for ADRs of 1.8 (range = 1–6, SD = 1.2). A total of 200 ADRs were identified in these patients, all but two of which were related to chemotherapy. The two cases concerned osteoporosis in patients receiving high-dose dexamethasone for multiple myeloma. The most frequent ADRs were aplasia or agranulocytosis (cited 156 times, 78 % of ADRs), which occurred in 57 patients (and led to re-admission for 41 patients) and polyneuropathy (24 times, 12 % of ADRs) in 12 patients (leading to re-admission for 5 patients). Group 2—Patients hospitalized for ADRs associated with other types of drugs at their first admission (see the supplementary material Table 2): there were 18 patients in this category, with a mean age of 77.2 years (range = 66– 88, SD = 6.5 years), a sex ratio of 1.0, and a mean number of re-admissions for ADR of 1.4 (range = 1–8, SD = 1.7). A total of 59 ADRs were identified in these patients. About one third of the ADRs were bleeding events inducing gastrointestinal ones (n = 10), hematoma (n = 3), anemia (n = 2), or epistaxis (n = 1). In nine cases, the bleeding ADRs were associated with an overdose of vitamin K antagonist (VKA) (Table 1). In ten cases, the same drug and ADR were suspected as the cause of the first and second hospital admissions (patients 1, 2, 4, 6, 8, 10, 11, 13, 14, 16, and 17). Three of these cases involved VKA overdoses (patients 4, 6, 8). Bleeding due to anticoagulants led to eight hospital admissions for patient 6. In total, 52 drugs were suspected of involvement in ADRs. Most were prescribed for chronic diseases (87 %). The initial prescription (known for 45 drugs) was made in hospital for 30 of these drugs and in an outpatient setting for the other 15. The suspected drugs were withdrawn in eight cases (48 %). Anticoagulants (VKA = 12, heparins = 7) were the drugs most frequently suspected, followed by psychotropic drugs (n = 6). Hospital re-admission was considered Bavoidable^ in four cases (22.2 %). In patient 1, amiodarone-induced hypothyroidism was diagnosed
Eur J Clin Pharmacol (2016) 72:631–639
635
Table 1 Comparison of characteristics of patients re-admitted for ADRs to those readmitted for other causes to hospital within 1 year of discharge from index admission Characteristics of patients during the first hospitalization
Rehospitalization with or without an ICD-10 code related to ADR
Univariate model
Multivariate model
No N = 410
Yes N = 87
p value
OR
95 % CI
p value
0.12
1
–
–
0.69
0.43–1.11
1 0.77
– 0.42–1.41
OR
95 % CI
Sex Male
203(49 %)
51(59 %)
Female Age (years)
207(51 %)
36(41 %)
[65–70] [71–75]
86(21 %) 84(20 %)
32(37 %) 24(27 %)
0.0009
–
–
–
–
–
– –
– –
113(28 %)
19(22 %)
0.45
0.45 0.24–0.85
–
–
127(31 %)
12(14 %)
0.25
0.12–0.52
–
–
246(60 %) 164(40 %)
47(54 %) 40(46 %)
0.3
1 1.28
– 0.80–2.03
–
– –
– –
161(39 %) 249(61 %)
14(16 %) 73(84 %)
80 Delay of the first hospitalization (days) < =7 >7 Score of Charlson comorbidity indexa 0—none > 0—at least one Diabetes
Congestive heart failure No Yes Kidney disease No Yes Metastatic cancer No Yes Lung disease No Yes Stroke No Yes Dementia No Yes
1
–
0.33
0.04–2.54
–
–
-
–
0.43
1 1.6
– 0.50–5.08
–
– –
– –
25(29 %) 62(71 %)
< 0.0001
1 7.69
– 4.59–12.88
PHARMACOEPIDEMIOLOGY AND PRESCRIPTION
Hospital re-admission associated with adverse drug reactions in patients over the age of 65 years Laurent Hauviller 1 & Frédéric Eyvrard 1 & Valérie Garnault 2 & Vanessa Rousseau 1 & L. Molinier 2 & Jean Louis Montastruc 1 & Haleh Bagheri 1
Received: 12 October 2015 / Accepted: 8 February 2016 / Published online: 17 February 2016 # Springer-Verlag Berlin Heidelberg 2016
Abstract Context Adverse drug reactions (ADRs) are responsible for 5 % of hospital admissions, but hospital re-admission induced by ADRs remains poorly documented. Objective The aim of this study was to estimate the rate of hospital re-admission and the factors associated with readmission in the patients over the age of 65 years. Secondary, we described the characteristics of cases of ADRs leading to re-admission for drugs other than chemotherapy agents. Methods Data were extracted from hospital discharge summaries provided by the Department of Medical Information of Toulouse University Hospital. All patients over the age of 65 years admitted to the hospital in 2010 for an ADR, identified from ICD-10 codes, were selected. All subsequent admissions of members of this cohort within 1 year of discharge following the index admission were reviewed retrospectively. The risk factors associated with hospital re-admission for ADRs were analyzed. Medical records were used for descriptive analysis of re-admission due to drugs other than chemotherapy agents.
Results We found that 553 of the 1000 patients admitted for ADRs in 2010 were re-admitted to hospital within 1 year. Among them, 87 cases were re-admitted for ADRs (estimated rate of 87/1000 re-admission for an ADR within 1 year). A comparison of the patients re-admitted for ADRs (n = 87) with those of patients re-admitted for other causes (n = 410) suggested that only cancer increased the risk of re-admission for ADRs (OR = 7.69 [4.59–12.88] 95 % CI). ADRs due to the same drug combination were the suspected cause of repeat admission in half the cases (other than chemotherapy). Hospital re-admission was considered avoidable in four cases (22 %). Conclusion This study shows an estimated rate of readmission for an ADR around 87/1000 within 1 year, and the same drug combination were the suspected cause of repeat admission in half the cases. At least, 11 % of cases were avoidable.
Electronic supplementary material The online version of this article (doi:10.1007/s00228-016-2022-4) contains supplementary material, which is available to authorized users.
Introduction
* Haleh Bagheri [email protected]
1
Service de Pharmacologie Médicale et Clinique, Centre Midi-Pyrénées de Pharmacovigilance, de Pharmacoépidémiologie et d’Informations sur le Médicament, Faculté de Médecine de l’Université Paul Sabatier-, INSERM U1027, 37 Allées Jules Guesde, 31000 Toulouse, France
2
Département d’Informations Médicales, CHU Toulouse, Hôtel Dieu St Jacques, 2 Rue Viguerie, 31300 Toulouse, France
Keywords Adverse drug reactions . Pharmacovigilance . Hospital re-admission . Avoidability
In western countries, several studies have investigated rates of hospital admission due to adverse drug reactions (ADRs). These studies have reported a prevalence of ADRs leading to hospitalization of 0.16 to 15.7 %, with an overall median prevalence of 5.3 % [1–3]. The large variation depends on the method of the study (retrospective or prospective) and the characteristics of included patients. In a recent review, Bouvy et al. suggested that the median rate of hospital admissions for ADRs was 3.5 % [4]. A higher rate would be expected in cancer patients (about 12 %), due to comorbidity and the toxicity of chemotherapy drugs [5, 6]. Likewise, the elderly
632
(people over the age of 65 years) are the largest consumers of healthcare and the recipients of more than half the drugs prescribed. It is therefore unsurprising that older people suffer more ADRs and are seven times more likely to require hospitalization due to an ADR. The incidence of ADRs leading to hospital admission increases to 9.8 % for those over the age of 75 years [2, 7–9]. Some of these ADRs leading to hospital admission should be avoidable. Some studies estimated that as many as 30–40 % of cases might have been avoided [9, 10]. Despite the major impact of hospital re-admission for ADRs on patient care and the burden placed on already overstretched hospitals, little is known about rates of re-admission to hospital due to ADRs. Two studies have suggested that between one fifth and one third of all re-admissions to hospital may be due to an ADR [8, 11]. Zhang suggested that the male gender, the duration of initial hospital stay, and Charlson comorbidity score (>7) were risk factors for repeat admission for ADRs, whereas Davies found that age increased the risk of re-admission for ADRS [8, 11]. In France, Bonnet-Zamponi et al. studied the impact of interventions designed to decrease hospital re-admission rates for patients aged 70 or over during a 6-month period following their discharge. They reported that 37.9 % of re-admissions were due to ADRs, but ADRs were not necessarily the cause of the initial hospitalization of these patients [12]. Taken into account, the scarcity of data, in 2010, we conducted a feasibility study in two units (a geriatric ward and an emergency ward) of Toulouse University Hospital over a period of 6 months to estimate the rate of hospital re-admission for ADRs in patients aged 65 and over. The first data suggested that 60 of the 556 patients admitted for ADRs in the first half of 2010 were re-admitted within 6 months of discharge, and an ADR was the reason for readmission in six of these patients [13]. We then undertook a larger study and extended our previous one to all units of Toulouse University Hospital, for a longer period (1 year of inclusion in 2010 and 1 year of follow-up of patients after discharge), to assess more accurately the rate of hospital re-admission and the factors associated with re-admission in the same population (patients aged >65 years). We also assessed the characteristics of the ADRs leading to repeat admissions, through a descriptive analysis. For this purpose, we choose to restrict the analysis to all readmissions associated with drugs other than the antineoplastic agents administered for cancer since it is well known that these drugs are involved in repeat hospital admissions and that an effective alternative treatment is lacking [5, 6].
Methods Design We carried out a retrospective cohort study at Toulouse University Hospital (2880 beds, in South West France).
Eur J Clin Pharmacol (2016) 72:631–639
Setting and population The data were obtained from the French hospital discharge database (Programme de Medicalisation des Systèmes d’Information, PMSI) and were provided by the Department of Medical Information of the hospital. Individual hospitalizations are entered into this database in the form of standardized hospital discharge summaries. In recent years, the PMSI and data from the national health insurance system have been linked, providing a useful tool for estimating the association between drug exposure and ADRs, e.g., valvulopathy due to benfluorex or thrombotic events due to the new generation of oral contraceptives [14, 15]. The data recorded include basic patient characteristics (gender, date of birth, and permanent patient identification number), days of admission and discharge, and the main and all related diagnoses. Diagnoses were encoded according to International Classification of Diseases, 10th revision (ICD-10), including all codes related to drug-induced ADRs, as used in our previous study [16, 17]. We used the definition of an ADR proposed by Edwards and Aronson [18]. The inclusion criteria were all admissions to Toulouse University Hospital of patients over the age of 65 years, over a period of 1 year (January 1–December 31, 2010), for which the discharge summary included a relevant ICD-10 code as the main or related diagnosis (sample A). We then used the permanent identification numbers of the patients selected in this first round of screening to identify all other re-admissions to Toulouse University Hospital for these patients within 1 year of discharge after the index admission (over a 1-year period from day 1 of release after the first hospitalization) (sample B, see Fig. 1). From this sample, we then selected the patients with relevant ICD-10 terms for the main or related diagnosis on the discharge summary for the repeat admission (sample C). Finally, we ensured that the ADR was correctly identified and was the cause of both initial admission and re-admission to hospital (rather than occurring during hospitalization), by checking the data for the patients in sample C, in the medical reports available via the hospital’s medical software (called ORBIS). We excluded cases for which the data noted in these medical records were insufficient for the assessment of ADRs corresponding to relevant ICD-10 terms and leading to hospital admission.
Data analysis In accordance with published findings, we investigated the association with ADR-related hospital re-admission of the following factors [7, 11]: patient age and gender, duration of the first hospitalization, and comorbidity according to the Charlson index as estimated from ICD-10 codes, for the main or related diagnosis of the corresponding hospital admission [19]. Given the risk of repeat admission associated with chemotherapy-induced ADRs, we analyzed the characteristics
Eur J Clin Pharmacol (2016) 72:631–639
Fig. 1 Selection of patients in a French hospital discharge database (Programme de Medicalisation des Systèmes d’Information, PMSI) using ICD-10 code related to ADRs during index admission and those re-admitted within 1 year of discharge from index admission
of patients re-admitted for chemotherapy-induced ADRs separately from those of patients re-admitted for ADRs associated with other types of drugs. Secondary, we carried out a descriptive analysis of cases of re-admission due to ADRs relating to drugs other than chemotherapy agents used for cancer treatment. For each case, we collected the following data: demographic data (age, gender), number of hospital admissions corresponding to ADRs (relevant ICD-10 code), and the suspected drug according to the anatomical-therapeutic classification [20], and hospital management (withdrawal or not of the suspected drug) and its pattern of prescription (prescription in an outpatient or hospital setting).
633
Given the retrospective nature of data collection in this study and the impossibility of knowing the time between drug intake and the occurrence of the ADR, we did not assess the causal relationship between ADRs and the suspected drug according to the French causality assessment method [21]. This method includes five scores (excluded, doubtful, possible, probable, certain) based on chronological (delay of occurrence of ADR, rechallenge) and semeiological (other explanations for ADRs) data. A preventability score was used for the suspected ADRs, according to Olivier’s scale with four categories of preventability of ADRs (Bpreventable,^ Bpotentially preventable,^ Bunclassable,^ Bnot preventable^), depending on data related to patient (risk factors), drug administration (error, interactions…), and possible alternative [22]. For the same reasons, we decided to focus only on cases for which the score obtained suggested that the ADR was Bpreventable.^ Statistical analysis was performed with SAS version 9.3 (SAS Institute, Cary, NC). A descriptive analysis was first performed on the characteristics of patients admitted or readmitted for ADRs. Age at the first hospital admission was analyzed both as a continuous variable and as a categorical variable: [65–70], [71–75], [76–80], and more than 80 years. The duration of the first hospitalization for ADR was classified into two categories: 7 days or less and more than 7 days as suggested by Davies [11]. Charlson index, estimated from the ICD-10 codes for the main or a related diagnosis for the corresponding hospital admission [19], was studied first as a binary variable: patients without comorbidity and patients with at least one comorbid condition. We then focused on the most frequent comorbid conditions in these patients: cancer, diabetes, heart failure, or kidney failure. We carried out a multivariate logistic regression analysis for categorical variables to compare the characteristics of patients re-admitted for ADRs with those re-admitted for other reasons. We began by inserting the variables with p values below 0.25 in univariate analyses. A descending, stepwise method was then used to exclude variables with p values below 0.05 from the final model. A survival analysis (Kaplan-Meier method) was performed to take into account the time between two hospital admissions in the two groups of patients: those re-admitted for ADRs and those re-admitted for other causes. This interval was defined as the time between the end of the first hospitalization, with the indication of an ICD-10 code relating to ADRs on the discharge report and the beginning of the next hospitalization. If the Bevent^ considered was re-admission for an ADR, then an absence of such readmission for an ADR within 1 year of the initial discharge was considered to be a Bnon-event.^ Logistic regression analysis was used to identify the risk factors associated with re-admission for an ADR (with an ICD-10 code relating to ADRs) after the initial hospitalization.
634
Results General data During 1 year from January 1 to December 31, 2010, data extraction from PMSI led to the identification of 1000 patients with at least one ICD-10 code relating to ADRs. The mean age of these patients was 77.6 years (range 65–104, SD = 7.4) and 511 (51.5 %) were female. In total, 1331 ICD-10 codes relating to ADRs were found in our sample of 1000 patients, mostly as related diagnoses (n = 1171, 88 %). The four most common codes were druginduced aplastic anemia (D61.1, n = 208, 15.6 %), unspecified adverse effect of drug (T88.7, n = 200, 15.0 %), adverse effects of anticoagulants (Y44.2, n = 171, 12.8 %), and agranulocytosis (D70, n = 72 5.4 %). A single ICD-10 code relating to ADR was found in the records of 812 patients, two codes were found for 122 patients and three were found for 32 others (34 patients had more than three ICD-10 codes relating to ADRs). About half the patients (n = 553) were re-admitted within 1 year of discharge following the index admission, for several causes. An ICD-10 code relating to ADRs was recorded for 143 of these patients. However, an analysis of the medical reports of these 143 patients led to exclude 56 cases for the following reasons: medical reports unavailable from the medical software for seven patients, absence of description of the ADRs corresponding to the ICD-10 code in the summary discharge letter for eight patients, or, in 41 patients, the ADR related to relevant ICD-10 code was present in the patient’s medical history but was not responsible for the most recent hospital admission (e.g., history of osteoporosis induced by corticosteroids). Analysis of factors associated with re-admission for ADRs A comparison of the characteristics (Table 1) of patients readmitted for ADRs (n = 87) with those of patients re-admitted for other causes (n = 410) found that only cancer increased the risk of re-admission for ADRs (OR = 7.69 [4.59–12.88]). No significant difference was found between the two groups for gender or duration of the first hospitalization (days). However, the mean age (years) of patients re-admitted for ADRs (74.0, SD = 6, range 65–89) was significantly lower than that of patients re-admitted for other causes (77.5, SD = 7.4, range 66–104) (Z test, p < 0.0001). The mean Charlson comorbidity score was higher in patients re-admitted for ADRs (5.6, SD = 1.9, range = 2–10) than in those re-admitted for other causes (4.6, SD = 1.9, range = 2–12) (Z test, p < 0.0004). Characteristics of the patients re-admitted for ADRs In total, 87 patients according to the inclusion criteria were readmitted for ADRs. According to the survival curve (see Supplementary Material Table 2), the probability of re-
Eur J Clin Pharmacol (2016) 72:631–639
admission for an ADR increased to 15 % within 4 months, thereafter remaining constant, at about 18 % (from 4 to 12 months). Eleven of the patients re-admitted for ADRs died during their hospital stay. Eight of these patients suffered from agranulocytosis due to chemotherapy. As described in the method, we analyzed the characteristics of patients hospitalized for ADRs associated with chemotherapy at the first admission (group 1) separately from those of patients hospitalized for ADRs associated with other types of drugs at their first admission (group 2): &
&
Group 1—Patients hospitalized for ADRs associated with chemotherapy at the first hospital admission: there were 69 patients in this category, with a mean age of 73.1 years (range = 65–89, SD = 5.7), a sex ratio of 1.5, and a mean number of re-admissions for ADRs of 1.8 (range = 1–6, SD = 1.2). A total of 200 ADRs were identified in these patients, all but two of which were related to chemotherapy. The two cases concerned osteoporosis in patients receiving high-dose dexamethasone for multiple myeloma. The most frequent ADRs were aplasia or agranulocytosis (cited 156 times, 78 % of ADRs), which occurred in 57 patients (and led to re-admission for 41 patients) and polyneuropathy (24 times, 12 % of ADRs) in 12 patients (leading to re-admission for 5 patients). Group 2—Patients hospitalized for ADRs associated with other types of drugs at their first admission (see the supplementary material Table 2): there were 18 patients in this category, with a mean age of 77.2 years (range = 66– 88, SD = 6.5 years), a sex ratio of 1.0, and a mean number of re-admissions for ADR of 1.4 (range = 1–8, SD = 1.7). A total of 59 ADRs were identified in these patients. About one third of the ADRs were bleeding events inducing gastrointestinal ones (n = 10), hematoma (n = 3), anemia (n = 2), or epistaxis (n = 1). In nine cases, the bleeding ADRs were associated with an overdose of vitamin K antagonist (VKA) (Table 1). In ten cases, the same drug and ADR were suspected as the cause of the first and second hospital admissions (patients 1, 2, 4, 6, 8, 10, 11, 13, 14, 16, and 17). Three of these cases involved VKA overdoses (patients 4, 6, 8). Bleeding due to anticoagulants led to eight hospital admissions for patient 6. In total, 52 drugs were suspected of involvement in ADRs. Most were prescribed for chronic diseases (87 %). The initial prescription (known for 45 drugs) was made in hospital for 30 of these drugs and in an outpatient setting for the other 15. The suspected drugs were withdrawn in eight cases (48 %). Anticoagulants (VKA = 12, heparins = 7) were the drugs most frequently suspected, followed by psychotropic drugs (n = 6). Hospital re-admission was considered Bavoidable^ in four cases (22.2 %). In patient 1, amiodarone-induced hypothyroidism was diagnosed
Eur J Clin Pharmacol (2016) 72:631–639
635
Table 1 Comparison of characteristics of patients re-admitted for ADRs to those readmitted for other causes to hospital within 1 year of discharge from index admission Characteristics of patients during the first hospitalization
Rehospitalization with or without an ICD-10 code related to ADR
Univariate model
Multivariate model
No N = 410
Yes N = 87
p value
OR
95 % CI
p value
0.12
1
–
–
0.69
0.43–1.11
1 0.77
– 0.42–1.41
OR
95 % CI
Sex Male
203(49 %)
51(59 %)
Female Age (years)
207(51 %)
36(41 %)
[65–70] [71–75]
86(21 %) 84(20 %)
32(37 %) 24(27 %)
0.0009
–
–
–
–
–
– –
– –
113(28 %)
19(22 %)
0.45
0.45 0.24–0.85
–
–
127(31 %)
12(14 %)
0.25
0.12–0.52
–
–
246(60 %) 164(40 %)
47(54 %) 40(46 %)
0.3
1 1.28
– 0.80–2.03
–
– –
– –
161(39 %) 249(61 %)
14(16 %) 73(84 %)
80 Delay of the first hospitalization (days) < =7 >7 Score of Charlson comorbidity indexa 0—none > 0—at least one Diabetes
Congestive heart failure No Yes Kidney disease No Yes Metastatic cancer No Yes Lung disease No Yes Stroke No Yes Dementia No Yes
1
–
0.33
0.04–2.54
–
–
-
–
0.43
1 1.6
– 0.50–5.08
–
– –
– –
25(29 %) 62(71 %)
< 0.0001
1 7.69
– 4.59–12.88
