American Journal of Infection Control 43 (2015) 839-43
Contents lists available at ScienceDirect
American Journal of Infection Control
American Journal of Infection Control
journal homepage: www.ajicjournal.org
Major article
Building-level analyses to prospectively detect influenza outbreaks in long-term care facilities: New York City, 2013-2014 Alison Levin-Rector MPH *, Beth Nivin MPH, Alice Yeung MPH, Annie D. Fine MD, Sharon K. Greene PhD, MPH Bureau of Communicable Disease, New York City Department of Health and Mental Hygiene, Queens, NY
Key Words: Disease cluster detection Geocoding Surveillance Automated analysis Nosocomial outbreaks Outbreak management
Background: Timely outbreak detection is necessary to successfully control influenza in long-term care facilities (LTCFs) and other institutions. To supplement nosocomial outbreak reports, calls from infection control staff, and active laboratory surveillance, the New York City (NYC) Department of Health and Mental Hygiene implemented an automated building-level analysis to proactively identify LTCFs with laboratory-confirmed influenza activity. Methods: Geocoded addresses of LTCFs in NYC were compared with geocoded residential addresses for all case-patients with laboratory-confirmed influenza reported through passive surveillance. An automated daily analysis used the geocoded building identification number, approximate text matching, and key-word searches to identify influenza in residents of LTCFs for review and follow-up by surveillance coordinators. Our aim was to determine whether the building analysis improved prospective outbreak detection during the 2013-2014 influenza season. Results: Of 119 outbreaks identified in LTCFs, 109 (92%) were ever detected by the building analysis, and 55 (46%) were first detected by the building analysis. Of the 5,953 LTCF staff and residents who received antiviral prophylaxis during the 2013-2014 season, 929 (16%) were at LTCFs where outbreaks were initially detected by the building analysis. Conclusions: A novel building-level analysis improved influenza outbreak identification in LTCFs in NYC, prompting timely infection control measures. Copyright Ó 2015 by the Association for Professionals in Infection Control and Epidemiology, Inc. Published by Elsevier Inc. All rights reserved.
Influenza is a serious health concern among elderly populations, as an estimated 90% of deaths due to influenza infection occur in persons aged 65 years and older.1 Influenza can be rapidly transmitted within nursing homes and other chronic-care facilities, affecting individuals at high risk for complications.2 Although longterm care facilities (LTCFs) in New York State are required by law to
* Address correspondence to Alison Levin-Rector, MPH, New York City Department of Health and Mental Hygiene, 42-09 28th St, WS 6-145, Queens, NY 11101. E-mail address: [email protected] (A. Levin-Rector). ALR and SKG were supported by the Public Health Emergency Preparedness Cooperative Agreement (grant No. 5U90TP000546-03) from the Centers for Disease Control and Prevention. BN was supported by the Epidemiology and Laboratory Capacity Cooperative Agreement (grant No. 3U50CK000407-01S1) from the Centers for Disease Control and Prevention. The contents of this publication are solely the responsibility of the authors and do not necessarily represent the official views of the Centers for Disease Control and Prevention. AY and ADF were supported by New York City tax levy funds. Conflicts of interest: None to report.
provide free influenza vaccines to residents and employees, vaccine coverage for employees has been low in recent seasons,3 and vaccine effectiveness in elderly persons can be modest.4,5 When influenza occurs in institutional settings, timely detection is critical to successfully control outbreaks through chemoprophylaxis and other infection control measures.6-10 Adults were at high risk for severe influenza illness and complications during the 2013-2014 influenza season, which was characterized in New York City (NYC) by a first wave predominated by pH1N1 activity and a second wave predominated by influenza B activity.11,12 Historically, influenza surveillance coordinators at the New York City Department of Health and Mental Hygiene (DOHMH) typically were first notified of influenza activity in LTCFs when the facility submitted a nosocomial outbreak report form to the New York State Department of Health (NYSDOH). In the 2010-2011 influenza season, to more proactively detect outbreaks, DOHMH initiated an analysis using approximate text matching and a key-word search on addresses to flag passive reports of laboratory-confirmed
0196-6553/$36.00 - Copyright Ó 2015 by the Association for Professionals in Infection Control and Epidemiology, Inc. Published by Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.ajic.2015.03.037
840
A. Levin-Rector et al. / American Journal of Infection Control 43 (2015) 839-43
influenza case-patients residing in LTCFs. This analysis seemed to identify outbreaks that may have otherwise gone unreported, but a formal evaluation was not performed. At the beginning of the 2013-2014 influenza season, an enhanced, automated building-level analysis was implemented to leverage routine geocoding of case-patients with laboratoryconfirmed influenza. Our objective was to evaluate the utility of this building analysis in improving the prospective detection of influenza outbreaks in LTCFs in NYC. MATERIALS AND METHODS Passive surveillance and geocoding of laboratory-confirmed influenza case-patients Cases of laboratory-confirmed influenza have been reportable in NYC since 2006.13,14 All clinical and commercial laboratories that perform testing report positive results of influenza tests (eg, rapid antigen test, viral culture, nucleic acid amplification test/polymerase chain reation, reverse-transcription polymerase chain reaction, and immunofluorescence) electronically via the New York State Electronic Clinical Laboratory Reporting System. Since mid-2012, each report received for NYC residents, whether electronically or through manual data entry, is processed (including using Coding Accuracy Support System address reference files from the US Postal Service to correct and standardize addresses15) and flows into a disease database system (Maven, Consilience Software, Austin, Tex). The disease database system is fully integrated with a geocoder using the LION geodatabase from the NYC Department of City Planning.16 The first geocodable residential address received in the disease database system for each case-patient is used to assign geocoded attributes such as latitude, longitude, and a building identification number (BIN). A BIN is an immutable 7-digit numerical identifier assigned by the Department of City Planning and unique to each building in NYC.17 BINs have previously been used for NYC surveillance purposes to track pesticide health effects18 and the safety of buildings following the World Trade Center attacks in 2001,19 and BINs have been included in routine geocoding output for cases of laboratory-confirmed influenza and other reportable diseases since mid-2012. Identifying influenza in LTCFs Eligible facilities In New York State, LTCFs are a type of Article 28 facility; that is, a health care facility that operates with a certificate granted under Article 28 of Public Health Law and is regulated by NYSDOH. Article 28 facilities are required to report the presence of confirmed or suspected influenza outbreaks (defined as 2 or more cases of influenza-like illness on the same unit within 7 days, or 1 laboratory-confirmed influenza case) to NYSDOH. For LTCFs in NYC, outbreak notification triggers a DOHMH investigation to characterize and monitor the outbreak and to provide guidance regarding appropriate infection control measures. Outbreak identification methods in prior use Article 28 facilities are required to report confirmed or suspected influenza outbreaks via the Nosocomial Outbreak Reporting Application (NORA), an online secure Web program.20 Once a NORA report is submitted and processed by NYSDOH, DOHMH influenza surveillance coordinators are alerted by e-mail. When LTCF patients are transferred to a hospital and test positive for influenza within 48 hours, hospital infection control staff often notify DOHMH of these nosocomial cases acquired in the
LTCF.21 In such instances when the affected LTCF staff are not also notified of the positive influenza test for their resident, a NORA report may not be submitted immediately. In addition, DOHMH conducts active laboratory surveillance during each influenza season. Commercial and hospital laboratories are contacted each week to obtain the number of respiratory specimens submitted for virologic testing and the number of positive specimens for influenza and other respiratory pathogens. Several of these are small commercial laboratories that conduct testing primarily for LTCFs and are more available than laboratories with a larger testing volume to collaborate with surveillance coordinators, facilitating active identification of positive cases in LTCF patients.22 Novel outbreak identification method At the beginning of the 2013-2014 influenza season, a list of the 175 nursing homes in the NY MetroeNYC region at the time was obtained from the NYSDOH Website.23 The addresses of these facilities were geocoded to obtain the BINs. We performed a building-level automated daily analysis to identify case-patients with laboratory-confirmed influenza in the disease database system whose address or BIN value matched that of a facility. The building analysis consisted of 4 steps: first, all addresses associated with influenza cases in the database system and all facility addresses went through a process to standardize street suffixes, numbers, and cardinal directions and delete ordinal suffixes, punctuation, and apartment or unit information. Second, we identified matching BIN values between the first geocodable residential address received for influenza case-patients in the database and LTCFs. Third, for addresses with missing or nonmatching BIN values, we performed an approximate text match between influenza casepatient addresses and LTCF addresses using the “compged” function in SAS 9.2 (SAS Institute, Inc, Cary, NC); sample SAS code for this step is provided in the Supplemental Online Appendix. We kept matches with a generalized edit distance, a measure of dissimilarity between 2 strings, of < 200 for manual review. We also required that the first 2 characters of both addresses must be identical. Midway through the 2013-2014 season, when a match was missed because the address in the disease database system included extra characters, we implemented a restriction to use for matching only the first 17 characters of the address text. The approximate textmatching process typically identified matches on case-patients whose addresses did not geocode (eg, because of an incorrect borough value or typographic error in the street name) and so were not assigned a BIN value but were similar to an address on the facility list. Finally, for addresses that were identified neither in the BIN match nor the approximate text match, we performed a keyword search to identify words like Home, Nursi, Center, or Care in any address field for a case-patient. Throughout the 2013-2014 influenza season (September 29, 2013-May 31, 201411), the SAS program described above was set up to run daily on the Task Scheduler (Microsoft, Redmond, Wash) of a computer dedicated to routine automated analyses. The code automatically flagged new matches and appended them to a master list in Excel 2003 (Microsoft). The first date that each match was discovered was recorded, and DOHMH influenza surveillance coordinators regularly inspected this master list of potential matches and decided whether further investigation was warranted for each record. As a secondary objective, we also performed this analysis for non-Article 28 congregate housing facilities (eg, adult care facilities and assisted living facilities) and homeless shelters. A list of nonArticle 28 facilities was obtained from the NYC Department of City Planning,24 and a list of homeless shelters was obtained from
A. Levin-Rector et al. / American Journal of Infection Control 43 (2015) 839-43
the NYC Department of Homeless Services (Dova Marder, personal communication). Unlike Article 28 facilities, these facilities are not legally required to report a single case of laboratory-confirmed influenza to NYSDOH. In contrast, these facilities are expected to comply with Article 11 of New York City’s Health Code, which requires immediate notification to DOHMH of any outbreak affecting 3 or more persons.13 Most of these facilities do not have medical providers or infection control staff onsite and are thus less equipped to manage outbreaks; given limited resources, timely identification of a single case of laboratory-confirmed influenza in these facilities was lower priority than in Article 28 facilities. When true matches indicating a case of laboratory-confirmed influenza were identified through manual review of building analysis results, surveillance coordinators contacted the appropriate facility or government agency. Control measures were implemented in response to all outbreaks as warranted, including prospective surveillance, cohorting ill residents, restricting floating, suspending group activities, reoffering vaccinations, instituting droplet precautions, excluding symptomatic staff, educating staff, limiting visitors or admissions, posting signs for visitors, and/or administering prophylaxis to staff and residents. Evaluation metrics To evaluate the utility of the building analysis, we calculated the percentage of all known outbreaks ever detected by the building analysis and the percentage of outbreaks detected by the building analysis before any other method, ie, NORA reports, hospital infection control notifications, or active laboratory surveillance. Of the outbreaks first detected by another method and later by the building analysis, we looked at the distribution in the number of days between first detection by each method. In addition, for all detected outbreaks, we tallied the total number of residents and staff placed on antiviral prophylaxis. Because this assessment was conducted in the course of routine public health practice, institutional review board approval was not required. RESULTS During the 2013-2014 influenza season, DOHMH received 15,876 reports for 13,508 laboratory-confirmed influenza cases among NYC residents. Of all reports, 15,744 (99%) were received through the Electronic Clinical Laboratory Reporting System. Of all cases, 13,442 (>99%) were reported through Electronic Clinical Laboratory Reporting System, and 12,681 (94%) were geocodable and assigned a BIN. The building analysis identified 426 matches between influenza case-patients in the disease database system and LTCFs. The median number of days between laboratory confirmation and match identification was 2 days (interquartile range, 2-4 days). Of these 426 matches, 234 (55%) were identified by BIN match, 188 (44%) were identified by approximate text match, and 4 (1%) were identified by key-word search. After manual review, 249 (58%) of these matches were determined to be true. The remaining 177 matches were deemed untrue either because approximate text matches were obviously not matches upon inspection, a patient address was quickly determined to be incorrectly reported, or a reported result from an initial rapid antigen test was determined to be a false positive after follow-up polymerase chain reaction testing. Of the 249 true matches, 223 (90%) were identified by BIN match, 22 (9%) were identified by approximate address match, and 4 (2%) were identified by key-word search. Upon further investigation, 12 of the true matches did not ultimately represent
841
Table 1 Influenza outbreaks in long-term care facilities in New York City, 2013-2014, by initial outbreak detection method
Detection method Building analysis Building analysis and another method on same day Nosocomial Outbreak Reporting Application* Hospital infection control staff Active laboratory surveillance Total
Number of outbreaks initially detected by specified method
Number (%) of outbreaks ever detected by building analysis
55 7
55 (100) 7 (100)
27
24 (89)
25 5 119
18 (72) 5 (100) 109 (92)
*An online secure Web program that long-term care facilities in New York State are required to use to report influenza outbreaks.
influenza activity in LTCFs, for example, because although the patient’s address was listed as an LTCF, the patient had actually been discharged long ago, or the patient actually had been hospitalized for > 48 hours and thus was unlikely to have been exposed in the LTCF. Of the 426 total matches, 237 true matches represented influenza activity in LTCFs (positive predictive value, 56%). These 237 true matches corresponded to 109 outbreaks within 70 LTCFs, out of 119 total known outbreaks within 75 LTCFs during the 2013-2014 influenza season. For context, during the 2001-2002 through 20122013 influenza seasons, the median number of outbreaks within LTCFs identified annually was 36 (range, 8-95). For 10 outbreaks that occurred in LTCFs during the 2013-2014 influenza season, the building analysis did not find any matches in the reportable disease database. Of the 119 outbreaks in 2013-2014, 109 were detected by the building analysis (sensitivity, 92%), and 55 outbreaks (46%) were detected by the building analysis before surveillance coordinators were notified through any other method. Of the remaining 64 outbreaks that were not initially identified by the building analysis, most (n ¼ 52; 81%) were initially detected by NORA report or hospital infection control staff, and most (n ¼ 54; 84%) were eventually also detected by the building analysis (Table 1). Of the 54 outbreaks that were initially detected by another method and later detected by the building analysis (n ¼ 47), or were detected by the building analysis and another method on the same day (n ¼ 7), the majority (n ¼ 37; 69%) were detected by the building analysis within 3 days of the date first reported through another method. In 6 instances, the building analysis identified a match longer than a week after it was initially reported. This lag was caused by various factors, such as delays in laboratory reporting, delays in processing laboratory reports with nonstandardized tests or results, or incomplete address information. One particularly long lag (33 days) was caused by a coding error. Across all 119 known outbreaks, 5,953 staff and residents received prophylaxis in the 2013-2014 season. Nine hundred twenty-nine (16%) of these people were at LTCFs where outbreaks were initially detected by the building analysis. For non-Article 28 facilities (ie, adult care facilities and assisted living facilities) and homeless shelters, the building analysis identified 166 and 224 matches, respectively. After manual review, 57 (34%) and 118 (53%) of these matches were determined to be true. Of these 175 true matches, 161 (92%) were identified by BIN match, and 14 (8%) were identified by approximate address matching. None of these matches were identified by key-word search. Because these facilities are not required to report individual cases of laboratory-confirmed influenza, it is unknown what percentage of all outbreaks in these facilities was detected by the building
842
A. Levin-Rector et al. / American Journal of Infection Control 43 (2015) 839-43
analysis, but facility administrators or the appropriate government agency were contacted and informed of matches. DISCUSSION Through routine geocoding of reportable disease data, an automated matching process facilitated identification of a record number of influenza outbreaks in LTCFs in NYC during the 20132014 influenza season, providing the opportunity for rapid institution of infection control measures and prophylaxis of several thousand staff and residents. Although the large number of influenza outbreaks may have been partially attributable to the severity of the circulating influenza virus strains, the automated building analysis also improved outbreak detection. For 37 of the 55 outbreaks that were initially detected by the building analysis, we did not rapidly receive subsequent reports by another method. Therefore, it was unknown whether reports were submitted because of prompting by DOHMH and would not have been submitted if DOHMH had not notified the facility. It is plausible that these 37 outbreaks would not have been detected at all if not for the building analysis. The building analysis has demonstrated utility for identifying influenza outbreaks in LTCFs, especially in situations when residents are transferred from an LTCF to a hospital and test positive for influenza, without the LTCF necessarily being aware of positive influenza test results for their residents. However, an important limitation is the requirement that address information in the reportable disease database is accurate. Missing addresses and address errors can cause cases to be missed or incorrectly assigned to LTCFs. In addition, the address associated with the report will only match if it corresponds with the LTCF that is the residence of the case-patient and not another address (eg, a prior address or a relative’s address). This method is also susceptible to reporting delays. Two of the 37 outbreaks that were only detected by the building analysis were identified too late for any public health action to be performed. Moreover, it is not possible to determine why the building analysis failed to detect 10 (8%) of the known outbreaks, because we do not know which case-patients in the disease database system, if any, corresponded to these 10 outbreaks. Reasons for failure of the building analysis in these instances could include LTCF residents not using the address of the facility as their home address, misspellings in the address field, errors in the geocoding process, the absence of laboratory-confirmed infection, failure to report, or a failure in transmission or processing of laboratory reports. For these reasons, the building analysis should complement and not replace existing reporting methods for LTCF outbreaks. The building analysis has applicability beyond influenza surveillance. We recently implemented similar analyses for other reportable diseases, specifically, detecting any reports in LTCF residents of amebiasis, cryptosporidiosis, cyclosporiasis, giardiasis, hepatitis A, acute hepatitis B, acute hepatitis C, legionellosis, listeriosis, respiratory syncytial virus, Shiga toxin-producing Escherichia coli, invasive group A Streptococcus, or invasive Streptococcus pneumoniae. These additional analyses have led to the successful detection of several disease events warranting investigation, including 2 listeriosis cases associated with a LTCF and a case of Legionnaires’ disease in an LTCF resident. Automated analyses using BIN and address matching also identify reports of certain diseases in other types of facilities, such as health care facilities, jails, and homeless shelters. The building analysis is not only useful for matching casepatient addresses with external lists of facilities but also for identifying 2 or more case-patients with the same residential or work address in any building who were diagnosed with the same disease
within a certain period of time. For example, this analysis has detected 2 cases of community-acquired Legionnaires’ disease in residents of the same NYC apartment building. For a health department to implement its own automated matching process to identify the occurrence of any reportable disease in any group of facilities, a list of facilities would first need to be obtained and geocoded. Whereas DOHMH uses a customized geocoding application, many external geocoders are available, including the ArcGIS Online Geocoding Service (ESRI, Redlands, Calif). Whereas these external services would not provide the unique BINs used in this analysis, other geocoding fields such as latitude and longitude would also be useful for detecting disease occurrence in certain locations corresponding to specific buildings. Geocoded elements can then be matched between facilities and case-patients with reportable diseases. In 2011, more than 80% of states had some level of electronic laboratory reporting, and in 2013, 61% routinely geocoded reportable disease data.25,26 If geocoding is not feasible, approximate address matching is a good alternative (see the Supplemental Online Appendix). The analytic program to conduct the matches can be set to run routinely on an automated scheduler, which we have found to be an especially low-cost and sustainable approach; after the initial setup, the time investment for the analyst is minimal and is limited to troubleshooting and enhancements. The program could also be run manually on an as-needed basis. The most time-intensive part of this process is unavoidably human review of potential matches and follow-up with facilities. We chose to be inclusive in our matching process, with only 58% of initial matches being deemed true and worthy of follow-up. If fewer resources are available, matching requirements can be adjusted to reduce false positive matches so that a more manageable number of potential matches are identified (eg, by decreasing the generalized edit distance threshold for text matching or by requiring that latitude and longitude values be exact matches). Most state health departments have at least the minimum data required to employ some level of case-patient-to-facility address matching. Automated analyses using routinely geocoded data should be considered to strengthen public health surveillance and response activities. Acknowledgments The authors thank Ana Maria Fireteanu, MPH, for providing data management and analytic support, Scott Harper, MD, MPH, for providing valuable feedback on this analysis, and the NYSDOH Healthcare Epidemiology and Infection Control Program. SUPPLEMENTARY DATA Supplementary data related to this article can be found at http:// dx.doi.org/10.1016/j.ajic.2015.03.037. References 1. Thompson WW, Shay DK, Weintraub E, Brammer L, Cox N, Anderson LJ, et al. Mortality associated with influenza and respiratory syncytial virus in the United States. JAMA 2003;289:179-86. 2. Menec VH, MacWilliam L, Aoki FY. Hospitalizations and deaths due to respiratory illnesses during influenza seasons: a comparison of community residents, senior housing residents, and nursing home residents. J Gerontol A Biol Sci Med Sci 2002;57:M629-35. 3. Person CJ, Nadeau JA, Schaffzin JK, et al. Influenza immunization coverage of residents and employees of long-term care facilities in New York State, 20002010. Am J Infect Control 2013;41:743-5. 4. Jefferson T, Rivetti D, Rivetti A, Rudin M, Di Pietrantonj C, Demicheli V. Efficacy and effectiveness of influenza vaccines in elderly people: a systematic review. Lancet 2005;366:1165-74.
A. Levin-Rector et al. / American Journal of Infection Control 43 (2015) 839-43 5. Simonsen L, Viboud C, Taylor RJ, Miller MA, Jackson L. Influenza vaccination and mortality benefits: new insights, new opportunities. Vaccine 2009;27: 6300-4. 6. Fiore AE, Fry A, Shay D, Gubareva L, Bresee JS, Uyeki TM. Antiviral agents for the treatment and chemoprophylaxis of influenza e recommendations of the Advisory Committee on Immunization Practices (ACIP). MMWR Recomm Rep 2011;60:1-24. 7. Zadeh MM, Buxton Bridges C, Thompson WW, Arden NH, Fukuda K. Influenza outbreak detection and control measures in nursing homes in the United States. J Am Geriatr Soc 2000;48:1310-5. 8. Rubin MS, Nivin B, Ackelsberg J. Effect of timing of amantadine chemoprophylaxis on severity of outbreaks of influenza A in adult long-term care facilities. Clin Infect Dis 2008;47:47-52. 9. Bamberg W, Finkner N, Guppy D, et al. Outbreaks of 2009 pandemic influenza A (H1N1) among long-term-care facility residents - three states, 2009. MMWR Morb Mortal Wkly Rep 2010;59:74-7. 10. Mahmud SM, Thompson LH, Nowicki DL, Plourde PJ. Outbreaks of influenzalike illness in long-term care facilities in Winnipeg, Canada. Influenza and Other Respiratory Viruses 2013;7:1055-61. 11. New York City Department of Health and Mental Hygiene. Influenza surveillance report: Week ending May 31, 2014 (week 22). Available from: http://www.nyc.gov/html/doh/flu/downloads/pdf/reports/weeklysurveillance05312014.pdf. Accessed July 18, 2014. 12. Epperson S, Blanton L, Kniss K, et al. Influenza activity - United States, 2013-14 season and composition of the 2014-15 influenza vaccines. MMWR Morb Mortal Wkly Rep 2014;63:483-90. 13. New York City Board of Health. Health Code and Rules. Article 11: Reportable diseases and conditions. Available from: http://www.nyc.gov/html/doh/ downloads/pdf/about/healthcode/health-code-article11.pdf. Accessed July 18, 2014. 14. New York City Board of Health. Notice of adoption of amendments to various sections of Article 11 of the New York City Health Code. 2005. Available from: http://www.nyc.gov/html/doh/downloads/pdf/public/ notice-adoption-article11.pdf. Accessed July 18, 2014.
843
15. United States Postal Service. CASS Technical Guide. Available from: https:// ribbs.usps.gov/cassmass/documents/tech_guides/. Accessed July 31, 2014. 16. New York City Department of City Planning. LION file geodatabase feature class. Available from: http://www.nyc.gov/html/dcp/html/bytes/applbyte. shtml#lion. Accessed July 18, 2014. 17. Department of City Planning, Information Technology Division, Geographic Systems Section. Geosupport Desktop Edition: User programming guide supplement, software version 10.1. 2007. 18. Kass DE, Thier AL, Leighton J, Cone JE, Jeffery NL. Developing a comprehensive pesticide health effects tracking system for an urban setting: New York City’s approach. Environ Health Perspect 2004;112:1419-23. 19. Kevany MJ. GIS in the World Trade Center attackdtrial by fire. Computers, Environment and Urban Systems 2003;27:517-83. 20. New York State Department of Health. Reporting requirements: NYSDOH Regulated Facilities (Article 28 Facilities), Nosocomial outbreak reporting. 2014. Available from: https://www.health.ny.gov/professionals/diseases/ reporting/communicable/infection/reporting.htm. Accessed August 5, 2014. 21. Nivin B, Yeung A. An unexpected benefit from enhanced 2009 H1N1 influenza surveillance. Am J Infect Control 2012;40:675-6. 22. Nivin B, Stoute A. Detection of nosocomial influenza outbreaks in long-term care facilities through active laboratory surveillance in New York City. Am J Infect Control 2008;36:498-9. 23. New York State Department of Health. New York State Nursing Home Profile. 2014. Available from: http://nursinghomes.nyhealth.gov/. Accessed September 1, 2013. 24. New York City Department of City Planning. Selected facilities and program sites in New York City: Data sources - release 2013. Available from: http://www.nyc. gov/html/dcp/html/bytes/applbyte.shtml#facilities2013. Accessed September 1, 2013. 25. Magnuson J. National electronic laboratory reporting (ELR) snapshot survey: summary of results. 2011. Available from: http://www.coast2coastinformatics. com/2011NationalELRSurvey-DataSummary.pdf. Accessed September 29, 2011. 26. Council of State and Territorial Epidemiologists (CSTE). 2013 National Assessment of Epidemiology Capacity. 2014. Available from: http://www.cste2.org/2013eca/ CSTEEpidemiologyCapacityAssessment2014-final2.pdf. Accessed March 10, 2015.
Contents lists available at ScienceDirect
American Journal of Infection Control
American Journal of Infection Control
journal homepage: www.ajicjournal.org
Major article
Building-level analyses to prospectively detect influenza outbreaks in long-term care facilities: New York City, 2013-2014 Alison Levin-Rector MPH *, Beth Nivin MPH, Alice Yeung MPH, Annie D. Fine MD, Sharon K. Greene PhD, MPH Bureau of Communicable Disease, New York City Department of Health and Mental Hygiene, Queens, NY
Key Words: Disease cluster detection Geocoding Surveillance Automated analysis Nosocomial outbreaks Outbreak management
Background: Timely outbreak detection is necessary to successfully control influenza in long-term care facilities (LTCFs) and other institutions. To supplement nosocomial outbreak reports, calls from infection control staff, and active laboratory surveillance, the New York City (NYC) Department of Health and Mental Hygiene implemented an automated building-level analysis to proactively identify LTCFs with laboratory-confirmed influenza activity. Methods: Geocoded addresses of LTCFs in NYC were compared with geocoded residential addresses for all case-patients with laboratory-confirmed influenza reported through passive surveillance. An automated daily analysis used the geocoded building identification number, approximate text matching, and key-word searches to identify influenza in residents of LTCFs for review and follow-up by surveillance coordinators. Our aim was to determine whether the building analysis improved prospective outbreak detection during the 2013-2014 influenza season. Results: Of 119 outbreaks identified in LTCFs, 109 (92%) were ever detected by the building analysis, and 55 (46%) were first detected by the building analysis. Of the 5,953 LTCF staff and residents who received antiviral prophylaxis during the 2013-2014 season, 929 (16%) were at LTCFs where outbreaks were initially detected by the building analysis. Conclusions: A novel building-level analysis improved influenza outbreak identification in LTCFs in NYC, prompting timely infection control measures. Copyright Ó 2015 by the Association for Professionals in Infection Control and Epidemiology, Inc. Published by Elsevier Inc. All rights reserved.
Influenza is a serious health concern among elderly populations, as an estimated 90% of deaths due to influenza infection occur in persons aged 65 years and older.1 Influenza can be rapidly transmitted within nursing homes and other chronic-care facilities, affecting individuals at high risk for complications.2 Although longterm care facilities (LTCFs) in New York State are required by law to
* Address correspondence to Alison Levin-Rector, MPH, New York City Department of Health and Mental Hygiene, 42-09 28th St, WS 6-145, Queens, NY 11101. E-mail address: [email protected] (A. Levin-Rector). ALR and SKG were supported by the Public Health Emergency Preparedness Cooperative Agreement (grant No. 5U90TP000546-03) from the Centers for Disease Control and Prevention. BN was supported by the Epidemiology and Laboratory Capacity Cooperative Agreement (grant No. 3U50CK000407-01S1) from the Centers for Disease Control and Prevention. The contents of this publication are solely the responsibility of the authors and do not necessarily represent the official views of the Centers for Disease Control and Prevention. AY and ADF were supported by New York City tax levy funds. Conflicts of interest: None to report.
provide free influenza vaccines to residents and employees, vaccine coverage for employees has been low in recent seasons,3 and vaccine effectiveness in elderly persons can be modest.4,5 When influenza occurs in institutional settings, timely detection is critical to successfully control outbreaks through chemoprophylaxis and other infection control measures.6-10 Adults were at high risk for severe influenza illness and complications during the 2013-2014 influenza season, which was characterized in New York City (NYC) by a first wave predominated by pH1N1 activity and a second wave predominated by influenza B activity.11,12 Historically, influenza surveillance coordinators at the New York City Department of Health and Mental Hygiene (DOHMH) typically were first notified of influenza activity in LTCFs when the facility submitted a nosocomial outbreak report form to the New York State Department of Health (NYSDOH). In the 2010-2011 influenza season, to more proactively detect outbreaks, DOHMH initiated an analysis using approximate text matching and a key-word search on addresses to flag passive reports of laboratory-confirmed
0196-6553/$36.00 - Copyright Ó 2015 by the Association for Professionals in Infection Control and Epidemiology, Inc. Published by Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.ajic.2015.03.037
840
A. Levin-Rector et al. / American Journal of Infection Control 43 (2015) 839-43
influenza case-patients residing in LTCFs. This analysis seemed to identify outbreaks that may have otherwise gone unreported, but a formal evaluation was not performed. At the beginning of the 2013-2014 influenza season, an enhanced, automated building-level analysis was implemented to leverage routine geocoding of case-patients with laboratoryconfirmed influenza. Our objective was to evaluate the utility of this building analysis in improving the prospective detection of influenza outbreaks in LTCFs in NYC. MATERIALS AND METHODS Passive surveillance and geocoding of laboratory-confirmed influenza case-patients Cases of laboratory-confirmed influenza have been reportable in NYC since 2006.13,14 All clinical and commercial laboratories that perform testing report positive results of influenza tests (eg, rapid antigen test, viral culture, nucleic acid amplification test/polymerase chain reation, reverse-transcription polymerase chain reaction, and immunofluorescence) electronically via the New York State Electronic Clinical Laboratory Reporting System. Since mid-2012, each report received for NYC residents, whether electronically or through manual data entry, is processed (including using Coding Accuracy Support System address reference files from the US Postal Service to correct and standardize addresses15) and flows into a disease database system (Maven, Consilience Software, Austin, Tex). The disease database system is fully integrated with a geocoder using the LION geodatabase from the NYC Department of City Planning.16 The first geocodable residential address received in the disease database system for each case-patient is used to assign geocoded attributes such as latitude, longitude, and a building identification number (BIN). A BIN is an immutable 7-digit numerical identifier assigned by the Department of City Planning and unique to each building in NYC.17 BINs have previously been used for NYC surveillance purposes to track pesticide health effects18 and the safety of buildings following the World Trade Center attacks in 2001,19 and BINs have been included in routine geocoding output for cases of laboratory-confirmed influenza and other reportable diseases since mid-2012. Identifying influenza in LTCFs Eligible facilities In New York State, LTCFs are a type of Article 28 facility; that is, a health care facility that operates with a certificate granted under Article 28 of Public Health Law and is regulated by NYSDOH. Article 28 facilities are required to report the presence of confirmed or suspected influenza outbreaks (defined as 2 or more cases of influenza-like illness on the same unit within 7 days, or 1 laboratory-confirmed influenza case) to NYSDOH. For LTCFs in NYC, outbreak notification triggers a DOHMH investigation to characterize and monitor the outbreak and to provide guidance regarding appropriate infection control measures. Outbreak identification methods in prior use Article 28 facilities are required to report confirmed or suspected influenza outbreaks via the Nosocomial Outbreak Reporting Application (NORA), an online secure Web program.20 Once a NORA report is submitted and processed by NYSDOH, DOHMH influenza surveillance coordinators are alerted by e-mail. When LTCF patients are transferred to a hospital and test positive for influenza within 48 hours, hospital infection control staff often notify DOHMH of these nosocomial cases acquired in the
LTCF.21 In such instances when the affected LTCF staff are not also notified of the positive influenza test for their resident, a NORA report may not be submitted immediately. In addition, DOHMH conducts active laboratory surveillance during each influenza season. Commercial and hospital laboratories are contacted each week to obtain the number of respiratory specimens submitted for virologic testing and the number of positive specimens for influenza and other respiratory pathogens. Several of these are small commercial laboratories that conduct testing primarily for LTCFs and are more available than laboratories with a larger testing volume to collaborate with surveillance coordinators, facilitating active identification of positive cases in LTCF patients.22 Novel outbreak identification method At the beginning of the 2013-2014 influenza season, a list of the 175 nursing homes in the NY MetroeNYC region at the time was obtained from the NYSDOH Website.23 The addresses of these facilities were geocoded to obtain the BINs. We performed a building-level automated daily analysis to identify case-patients with laboratory-confirmed influenza in the disease database system whose address or BIN value matched that of a facility. The building analysis consisted of 4 steps: first, all addresses associated with influenza cases in the database system and all facility addresses went through a process to standardize street suffixes, numbers, and cardinal directions and delete ordinal suffixes, punctuation, and apartment or unit information. Second, we identified matching BIN values between the first geocodable residential address received for influenza case-patients in the database and LTCFs. Third, for addresses with missing or nonmatching BIN values, we performed an approximate text match between influenza casepatient addresses and LTCF addresses using the “compged” function in SAS 9.2 (SAS Institute, Inc, Cary, NC); sample SAS code for this step is provided in the Supplemental Online Appendix. We kept matches with a generalized edit distance, a measure of dissimilarity between 2 strings, of < 200 for manual review. We also required that the first 2 characters of both addresses must be identical. Midway through the 2013-2014 season, when a match was missed because the address in the disease database system included extra characters, we implemented a restriction to use for matching only the first 17 characters of the address text. The approximate textmatching process typically identified matches on case-patients whose addresses did not geocode (eg, because of an incorrect borough value or typographic error in the street name) and so were not assigned a BIN value but were similar to an address on the facility list. Finally, for addresses that were identified neither in the BIN match nor the approximate text match, we performed a keyword search to identify words like Home, Nursi, Center, or Care in any address field for a case-patient. Throughout the 2013-2014 influenza season (September 29, 2013-May 31, 201411), the SAS program described above was set up to run daily on the Task Scheduler (Microsoft, Redmond, Wash) of a computer dedicated to routine automated analyses. The code automatically flagged new matches and appended them to a master list in Excel 2003 (Microsoft). The first date that each match was discovered was recorded, and DOHMH influenza surveillance coordinators regularly inspected this master list of potential matches and decided whether further investigation was warranted for each record. As a secondary objective, we also performed this analysis for non-Article 28 congregate housing facilities (eg, adult care facilities and assisted living facilities) and homeless shelters. A list of nonArticle 28 facilities was obtained from the NYC Department of City Planning,24 and a list of homeless shelters was obtained from
A. Levin-Rector et al. / American Journal of Infection Control 43 (2015) 839-43
the NYC Department of Homeless Services (Dova Marder, personal communication). Unlike Article 28 facilities, these facilities are not legally required to report a single case of laboratory-confirmed influenza to NYSDOH. In contrast, these facilities are expected to comply with Article 11 of New York City’s Health Code, which requires immediate notification to DOHMH of any outbreak affecting 3 or more persons.13 Most of these facilities do not have medical providers or infection control staff onsite and are thus less equipped to manage outbreaks; given limited resources, timely identification of a single case of laboratory-confirmed influenza in these facilities was lower priority than in Article 28 facilities. When true matches indicating a case of laboratory-confirmed influenza were identified through manual review of building analysis results, surveillance coordinators contacted the appropriate facility or government agency. Control measures were implemented in response to all outbreaks as warranted, including prospective surveillance, cohorting ill residents, restricting floating, suspending group activities, reoffering vaccinations, instituting droplet precautions, excluding symptomatic staff, educating staff, limiting visitors or admissions, posting signs for visitors, and/or administering prophylaxis to staff and residents. Evaluation metrics To evaluate the utility of the building analysis, we calculated the percentage of all known outbreaks ever detected by the building analysis and the percentage of outbreaks detected by the building analysis before any other method, ie, NORA reports, hospital infection control notifications, or active laboratory surveillance. Of the outbreaks first detected by another method and later by the building analysis, we looked at the distribution in the number of days between first detection by each method. In addition, for all detected outbreaks, we tallied the total number of residents and staff placed on antiviral prophylaxis. Because this assessment was conducted in the course of routine public health practice, institutional review board approval was not required. RESULTS During the 2013-2014 influenza season, DOHMH received 15,876 reports for 13,508 laboratory-confirmed influenza cases among NYC residents. Of all reports, 15,744 (99%) were received through the Electronic Clinical Laboratory Reporting System. Of all cases, 13,442 (>99%) were reported through Electronic Clinical Laboratory Reporting System, and 12,681 (94%) were geocodable and assigned a BIN. The building analysis identified 426 matches between influenza case-patients in the disease database system and LTCFs. The median number of days between laboratory confirmation and match identification was 2 days (interquartile range, 2-4 days). Of these 426 matches, 234 (55%) were identified by BIN match, 188 (44%) were identified by approximate text match, and 4 (1%) were identified by key-word search. After manual review, 249 (58%) of these matches were determined to be true. The remaining 177 matches were deemed untrue either because approximate text matches were obviously not matches upon inspection, a patient address was quickly determined to be incorrectly reported, or a reported result from an initial rapid antigen test was determined to be a false positive after follow-up polymerase chain reaction testing. Of the 249 true matches, 223 (90%) were identified by BIN match, 22 (9%) were identified by approximate address match, and 4 (2%) were identified by key-word search. Upon further investigation, 12 of the true matches did not ultimately represent
841
Table 1 Influenza outbreaks in long-term care facilities in New York City, 2013-2014, by initial outbreak detection method
Detection method Building analysis Building analysis and another method on same day Nosocomial Outbreak Reporting Application* Hospital infection control staff Active laboratory surveillance Total
Number of outbreaks initially detected by specified method
Number (%) of outbreaks ever detected by building analysis
55 7
55 (100) 7 (100)
27
24 (89)
25 5 119
18 (72) 5 (100) 109 (92)
*An online secure Web program that long-term care facilities in New York State are required to use to report influenza outbreaks.
influenza activity in LTCFs, for example, because although the patient’s address was listed as an LTCF, the patient had actually been discharged long ago, or the patient actually had been hospitalized for > 48 hours and thus was unlikely to have been exposed in the LTCF. Of the 426 total matches, 237 true matches represented influenza activity in LTCFs (positive predictive value, 56%). These 237 true matches corresponded to 109 outbreaks within 70 LTCFs, out of 119 total known outbreaks within 75 LTCFs during the 2013-2014 influenza season. For context, during the 2001-2002 through 20122013 influenza seasons, the median number of outbreaks within LTCFs identified annually was 36 (range, 8-95). For 10 outbreaks that occurred in LTCFs during the 2013-2014 influenza season, the building analysis did not find any matches in the reportable disease database. Of the 119 outbreaks in 2013-2014, 109 were detected by the building analysis (sensitivity, 92%), and 55 outbreaks (46%) were detected by the building analysis before surveillance coordinators were notified through any other method. Of the remaining 64 outbreaks that were not initially identified by the building analysis, most (n ¼ 52; 81%) were initially detected by NORA report or hospital infection control staff, and most (n ¼ 54; 84%) were eventually also detected by the building analysis (Table 1). Of the 54 outbreaks that were initially detected by another method and later detected by the building analysis (n ¼ 47), or were detected by the building analysis and another method on the same day (n ¼ 7), the majority (n ¼ 37; 69%) were detected by the building analysis within 3 days of the date first reported through another method. In 6 instances, the building analysis identified a match longer than a week after it was initially reported. This lag was caused by various factors, such as delays in laboratory reporting, delays in processing laboratory reports with nonstandardized tests or results, or incomplete address information. One particularly long lag (33 days) was caused by a coding error. Across all 119 known outbreaks, 5,953 staff and residents received prophylaxis in the 2013-2014 season. Nine hundred twenty-nine (16%) of these people were at LTCFs where outbreaks were initially detected by the building analysis. For non-Article 28 facilities (ie, adult care facilities and assisted living facilities) and homeless shelters, the building analysis identified 166 and 224 matches, respectively. After manual review, 57 (34%) and 118 (53%) of these matches were determined to be true. Of these 175 true matches, 161 (92%) were identified by BIN match, and 14 (8%) were identified by approximate address matching. None of these matches were identified by key-word search. Because these facilities are not required to report individual cases of laboratory-confirmed influenza, it is unknown what percentage of all outbreaks in these facilities was detected by the building
842
A. Levin-Rector et al. / American Journal of Infection Control 43 (2015) 839-43
analysis, but facility administrators or the appropriate government agency were contacted and informed of matches. DISCUSSION Through routine geocoding of reportable disease data, an automated matching process facilitated identification of a record number of influenza outbreaks in LTCFs in NYC during the 20132014 influenza season, providing the opportunity for rapid institution of infection control measures and prophylaxis of several thousand staff and residents. Although the large number of influenza outbreaks may have been partially attributable to the severity of the circulating influenza virus strains, the automated building analysis also improved outbreak detection. For 37 of the 55 outbreaks that were initially detected by the building analysis, we did not rapidly receive subsequent reports by another method. Therefore, it was unknown whether reports were submitted because of prompting by DOHMH and would not have been submitted if DOHMH had not notified the facility. It is plausible that these 37 outbreaks would not have been detected at all if not for the building analysis. The building analysis has demonstrated utility for identifying influenza outbreaks in LTCFs, especially in situations when residents are transferred from an LTCF to a hospital and test positive for influenza, without the LTCF necessarily being aware of positive influenza test results for their residents. However, an important limitation is the requirement that address information in the reportable disease database is accurate. Missing addresses and address errors can cause cases to be missed or incorrectly assigned to LTCFs. In addition, the address associated with the report will only match if it corresponds with the LTCF that is the residence of the case-patient and not another address (eg, a prior address or a relative’s address). This method is also susceptible to reporting delays. Two of the 37 outbreaks that were only detected by the building analysis were identified too late for any public health action to be performed. Moreover, it is not possible to determine why the building analysis failed to detect 10 (8%) of the known outbreaks, because we do not know which case-patients in the disease database system, if any, corresponded to these 10 outbreaks. Reasons for failure of the building analysis in these instances could include LTCF residents not using the address of the facility as their home address, misspellings in the address field, errors in the geocoding process, the absence of laboratory-confirmed infection, failure to report, or a failure in transmission or processing of laboratory reports. For these reasons, the building analysis should complement and not replace existing reporting methods for LTCF outbreaks. The building analysis has applicability beyond influenza surveillance. We recently implemented similar analyses for other reportable diseases, specifically, detecting any reports in LTCF residents of amebiasis, cryptosporidiosis, cyclosporiasis, giardiasis, hepatitis A, acute hepatitis B, acute hepatitis C, legionellosis, listeriosis, respiratory syncytial virus, Shiga toxin-producing Escherichia coli, invasive group A Streptococcus, or invasive Streptococcus pneumoniae. These additional analyses have led to the successful detection of several disease events warranting investigation, including 2 listeriosis cases associated with a LTCF and a case of Legionnaires’ disease in an LTCF resident. Automated analyses using BIN and address matching also identify reports of certain diseases in other types of facilities, such as health care facilities, jails, and homeless shelters. The building analysis is not only useful for matching casepatient addresses with external lists of facilities but also for identifying 2 or more case-patients with the same residential or work address in any building who were diagnosed with the same disease
within a certain period of time. For example, this analysis has detected 2 cases of community-acquired Legionnaires’ disease in residents of the same NYC apartment building. For a health department to implement its own automated matching process to identify the occurrence of any reportable disease in any group of facilities, a list of facilities would first need to be obtained and geocoded. Whereas DOHMH uses a customized geocoding application, many external geocoders are available, including the ArcGIS Online Geocoding Service (ESRI, Redlands, Calif). Whereas these external services would not provide the unique BINs used in this analysis, other geocoding fields such as latitude and longitude would also be useful for detecting disease occurrence in certain locations corresponding to specific buildings. Geocoded elements can then be matched between facilities and case-patients with reportable diseases. In 2011, more than 80% of states had some level of electronic laboratory reporting, and in 2013, 61% routinely geocoded reportable disease data.25,26 If geocoding is not feasible, approximate address matching is a good alternative (see the Supplemental Online Appendix). The analytic program to conduct the matches can be set to run routinely on an automated scheduler, which we have found to be an especially low-cost and sustainable approach; after the initial setup, the time investment for the analyst is minimal and is limited to troubleshooting and enhancements. The program could also be run manually on an as-needed basis. The most time-intensive part of this process is unavoidably human review of potential matches and follow-up with facilities. We chose to be inclusive in our matching process, with only 58% of initial matches being deemed true and worthy of follow-up. If fewer resources are available, matching requirements can be adjusted to reduce false positive matches so that a more manageable number of potential matches are identified (eg, by decreasing the generalized edit distance threshold for text matching or by requiring that latitude and longitude values be exact matches). Most state health departments have at least the minimum data required to employ some level of case-patient-to-facility address matching. Automated analyses using routinely geocoded data should be considered to strengthen public health surveillance and response activities. Acknowledgments The authors thank Ana Maria Fireteanu, MPH, for providing data management and analytic support, Scott Harper, MD, MPH, for providing valuable feedback on this analysis, and the NYSDOH Healthcare Epidemiology and Infection Control Program. SUPPLEMENTARY DATA Supplementary data related to this article can be found at http:// dx.doi.org/10.1016/j.ajic.2015.03.037. References 1. Thompson WW, Shay DK, Weintraub E, Brammer L, Cox N, Anderson LJ, et al. Mortality associated with influenza and respiratory syncytial virus in the United States. JAMA 2003;289:179-86. 2. Menec VH, MacWilliam L, Aoki FY. Hospitalizations and deaths due to respiratory illnesses during influenza seasons: a comparison of community residents, senior housing residents, and nursing home residents. J Gerontol A Biol Sci Med Sci 2002;57:M629-35. 3. Person CJ, Nadeau JA, Schaffzin JK, et al. Influenza immunization coverage of residents and employees of long-term care facilities in New York State, 20002010. Am J Infect Control 2013;41:743-5. 4. Jefferson T, Rivetti D, Rivetti A, Rudin M, Di Pietrantonj C, Demicheli V. Efficacy and effectiveness of influenza vaccines in elderly people: a systematic review. Lancet 2005;366:1165-74.
A. Levin-Rector et al. / American Journal of Infection Control 43 (2015) 839-43 5. Simonsen L, Viboud C, Taylor RJ, Miller MA, Jackson L. Influenza vaccination and mortality benefits: new insights, new opportunities. Vaccine 2009;27: 6300-4. 6. Fiore AE, Fry A, Shay D, Gubareva L, Bresee JS, Uyeki TM. Antiviral agents for the treatment and chemoprophylaxis of influenza e recommendations of the Advisory Committee on Immunization Practices (ACIP). MMWR Recomm Rep 2011;60:1-24. 7. Zadeh MM, Buxton Bridges C, Thompson WW, Arden NH, Fukuda K. Influenza outbreak detection and control measures in nursing homes in the United States. J Am Geriatr Soc 2000;48:1310-5. 8. Rubin MS, Nivin B, Ackelsberg J. Effect of timing of amantadine chemoprophylaxis on severity of outbreaks of influenza A in adult long-term care facilities. Clin Infect Dis 2008;47:47-52. 9. Bamberg W, Finkner N, Guppy D, et al. Outbreaks of 2009 pandemic influenza A (H1N1) among long-term-care facility residents - three states, 2009. MMWR Morb Mortal Wkly Rep 2010;59:74-7. 10. Mahmud SM, Thompson LH, Nowicki DL, Plourde PJ. Outbreaks of influenzalike illness in long-term care facilities in Winnipeg, Canada. Influenza and Other Respiratory Viruses 2013;7:1055-61. 11. New York City Department of Health and Mental Hygiene. Influenza surveillance report: Week ending May 31, 2014 (week 22). Available from: http://www.nyc.gov/html/doh/flu/downloads/pdf/reports/weeklysurveillance05312014.pdf. Accessed July 18, 2014. 12. Epperson S, Blanton L, Kniss K, et al. Influenza activity - United States, 2013-14 season and composition of the 2014-15 influenza vaccines. MMWR Morb Mortal Wkly Rep 2014;63:483-90. 13. New York City Board of Health. Health Code and Rules. Article 11: Reportable diseases and conditions. Available from: http://www.nyc.gov/html/doh/ downloads/pdf/about/healthcode/health-code-article11.pdf. Accessed July 18, 2014. 14. New York City Board of Health. Notice of adoption of amendments to various sections of Article 11 of the New York City Health Code. 2005. Available from: http://www.nyc.gov/html/doh/downloads/pdf/public/ notice-adoption-article11.pdf. Accessed July 18, 2014.
843
15. United States Postal Service. CASS Technical Guide. Available from: https:// ribbs.usps.gov/cassmass/documents/tech_guides/. Accessed July 31, 2014. 16. New York City Department of City Planning. LION file geodatabase feature class. Available from: http://www.nyc.gov/html/dcp/html/bytes/applbyte. shtml#lion. Accessed July 18, 2014. 17. Department of City Planning, Information Technology Division, Geographic Systems Section. Geosupport Desktop Edition: User programming guide supplement, software version 10.1. 2007. 18. Kass DE, Thier AL, Leighton J, Cone JE, Jeffery NL. Developing a comprehensive pesticide health effects tracking system for an urban setting: New York City’s approach. Environ Health Perspect 2004;112:1419-23. 19. Kevany MJ. GIS in the World Trade Center attackdtrial by fire. Computers, Environment and Urban Systems 2003;27:517-83. 20. New York State Department of Health. Reporting requirements: NYSDOH Regulated Facilities (Article 28 Facilities), Nosocomial outbreak reporting. 2014. Available from: https://www.health.ny.gov/professionals/diseases/ reporting/communicable/infection/reporting.htm. Accessed August 5, 2014. 21. Nivin B, Yeung A. An unexpected benefit from enhanced 2009 H1N1 influenza surveillance. Am J Infect Control 2012;40:675-6. 22. Nivin B, Stoute A. Detection of nosocomial influenza outbreaks in long-term care facilities through active laboratory surveillance in New York City. Am J Infect Control 2008;36:498-9. 23. New York State Department of Health. New York State Nursing Home Profile. 2014. Available from: http://nursinghomes.nyhealth.gov/. Accessed September 1, 2013. 24. New York City Department of City Planning. Selected facilities and program sites in New York City: Data sources - release 2013. Available from: http://www.nyc. gov/html/dcp/html/bytes/applbyte.shtml#facilities2013. Accessed September 1, 2013. 25. Magnuson J. National electronic laboratory reporting (ELR) snapshot survey: summary of results. 2011. Available from: http://www.coast2coastinformatics. com/2011NationalELRSurvey-DataSummary.pdf. Accessed September 29, 2011. 26. Council of State and Territorial Epidemiologists (CSTE). 2013 National Assessment of Epidemiology Capacity. 2014. Available from: http://www.cste2.org/2013eca/ CSTEEpidemiologyCapacityAssessment2014-final2.pdf. Accessed March 10, 2015.
