Facility A

In October 2012, during monthly medication cart inspections on inpatient units, single-patient use insulin pens with no patient labels were discovered in medication carts. During interviews with nursing staff, several nurses acknowledged reusing insulin pens on multiple patients but used a new sterile needle for each injection. Owing to lack of labeling, it was unclear how long pens had been in the cart or how many patients had been exposed to each pen. Therefore, anyone receiving glargine insulin by insulin pen (only insulin type given by pen in this inpatient setting) was considered at risk and included in our investigation. Insulin pen use at this facility was discontinued in December 2012. ¹

Facility B

In January 2013, a national VA review of insulin pen practices was initiated after identification of the Facility A incident. Despite having insulin pens labeled with individual patient information and bar-coded medication administration data, a licensed practical nurse working in a VA Community Living Center reported she reused insulin pens between Community Living Center residents. Additionally, a registered nurse was identified as having reused insulin pens between patients on an inpatient unit. Both nurses reported changing needles between patients but it was unclear how long this practice may have occurred. Therefore, anyone receiving insulin via pen in both Community Living Center and acute care units was considered at risk and included in our investigation. The facility took immediate action and discontinued use as of January 10, 2013. ¹²

Internal VA review board/subject matter expert panels determined that epidemiologic investigations were warranted. In February 2013, patients at risk from both facilities were identified, and a lookback investigation was initiated. At-risk patients included inpatients receiving glargine insulin from Lantus SoloSTAR pens from October 19, 2010, through December 31, 2012, at facility A and Community Living Center and acute care inpatients receiving insulin from Lantus SoloSTAR pens, Detemir Flexpens, and Novolog Flexpens from October 5, 2010, through January 10, 2013, at Facility B, and were identified using VA’s electronic health record and Corporate Data Warehouse bar-coded medication administration data. Date ranges for patient identification included start of insulin pen usage in these facilities and continued past the time of discontinuation to ensure all potential exposure patients were identified. Facilities notified patients of their possible exposure; the patients were then offered and verbally consented to BBP testing for HBV, HCV, and HIV. Patients whose potential exposure occurred less than 6 months before time of testing were requested to return for repeat testing. Laboratory and clinical data for exposed patients were reviewed to determine pre–insulin pen exposure serologic status for HBV, HCV, and HIV infection and confirmation of insulin receipt using an insulin pen. Records from all VA facilities visited by patients were reviewed, going back as far as electronic health record documentation was available along with available Department of Defense and civilian medical records in VA’s electronic health record. Since the goal was to assess BBP infection transmission risk among individuals exposed to reused insulin pens, it was considered an urgent public health investigation and did not require VA or facility institutional review board approval. ¹³

Case (“newly identified”) patients were defined as having positive post–insulin pen exposure serologic test results for HBV (hepatitis B surface antigen [HbsAg] or core total antibody [HbcAb] positive), HCV (HCV antibody [HCV Ab] positive with confirmatory antibody testing with alternative assays) or HIV (HIV-1 antibody [HIV Ab] and Western blot positive) infection, had been an inpatient at the facility during the defined period, and had either negative or no serologic testing documented in VA’s electronic health record prior to insulin pen exposure. Potential proximate (“previously infected”) patients were those with positive pre–insulin pen administration serologic test result(s) for 1 or more viruses available or who were known to be chronically infected with HBV, HCV, or HIV and received insulin via an insulin pen while an inpatient during the defined period. Case patients were not temporally matched with specific potential proximate patients since it was unclear how long unlabeled insulin pens had been in medication carts or which patients were exposed to which pens.

Viral serologic testing was performed at each facility and in some cases additional testing was conducted at VA’s Public Health Reference Laboratory in Palo Alto, California. HBV testing included HbsAg, HbcAb, hepatitis B surface antibody, and when indicated, HBV viral load testing. HCV testing included HCV Ab, and when initial screen was positive, a confirmatory antibody test with an alternative antibody assay, HCV viral load, and genotype testing. HIV testing included HIV 1/2 Ab, and when indicated, HIV-1 confirmatory Western blot and HIV-1 viral load (Online Table 1).

Background inpatient and outpatient seroprevalence for HIV, HBV, and HCV infection at each facility from January 1, 2010, through May 28, 2014, was obtained from the VA Healthcare Associated Infection and Influenza Surveillance System, which contains national VA patient infectious diseases–related laboratory results. Unique patients were used for calculation purposes and exposed cohort patients were included within the total number of unique patients tested. Fisher exact statistics comparing cohort and facility background seroprevalence were calculated under the null hypothesis that standardized morbidity ratios of observed (cohort) versus expected (based on facility) counts were 1 if there was no significant difference. Calculations were conducted using OpenEpi, version 3, and were considered significantly different at P<.05.Reference Sullivan, Dean and Soe ¹⁴

For patients with sufficient HCV viral load present, single-genome sequencing of the HCV envelope E1-E2 gene region (300 base pairs) was performed to assess relatedness of viral strains from patients with chronic and incident HCV infection.Reference Bukh, Miller and Purcell ^{15 ,} Reference Kato ¹⁶ Briefly, all laboratory investigations were performed using established methods and rigorous quality control procedures, as described previously.Reference Cody, Nainan and Garfein ^{17 –} Reference Thompson, Hellinger and Kay ²¹ Extracted plasma HCV RNA was reverse transcribed with random primers, diluted, and amplified by nested polymerase chain reaction. Following capillary electrophoresis analysis, only amplicons generated at cDNA dilutions where only 1 of every 3 reactions was positive (thus considered to arise from a single genome) were subsequently sequenced bidirectionally. During sequence assembly, any reactions showing evidence of mixtures at any nucleotide position were discarded. Pairwise analysis of nucleotide sequences (>20 per patient) and phylogenetic tree construction (using Geneious, version 7.1.3, software) were used to evaluate genetic relatedness between newly infected, proximal, and reference patient strains. Trees were created using Kimura 2-parameter distance matrix with 1,000 bootstrap replicates. Patient strains demonstrating less than 5% difference in nucleotide sequence were considered possibly related.