What we have done, with which data, by which methods.

Studies our researchers have presented and published, told the way you would assess one. They come from the public record, not from client engagements, and every figure is quoted as the authors reported it.

1. 01Pharmacovigilance

   A claims-based sentinel system for drug safety

   47,319

   indication–drug–outcome combinations, in under 48 hours

   Can relationships between indications, drugs, and outcomes be surfaced from routine claims fast enough to direct further study?

   Data, method, and finding＋

   Data

   De-identified claims for five million people over twelve months.

   Method

   Relative risk and chi-square for every combination of 73 drug categories and 39 health outcomes of interest. Indication from ICD-9 claims, exposure from NDC refills, and an outcome counted only if it appeared in months 6–12 and not in months 1–6.

   Finding

   The system produced 47,319 combinations in under 48 hours. At a threshold of at least five patients and a chi-square above 5, 13,094 of them (28%) were flagged for closer inspection — a way of targeting analysis rather than a list of conclusions.

   Wei H, Mehta R, Steinberg G. ISPOR 18th Annual International Meeting, 2013.

2. 02Infectious disease

   Incidence and cost of healthcare-associated infections

   $33,626

   median total cost per patient with a healthcare-associated infection

   What do device- and non-device-related infections actually cost an integrated health system, and how often do they occur?

   Data, method, and finding＋

   Data

   79,839 admissions from 62,647 adult patients at two urban hospitals, July 2008 to September 2010.

   Method

   Electronic surveillance with every case verified against NHSN definitions by three infection preventionists; clinical, cost, and demographic data drawn from the enterprise data warehouse.

   Finding

   2,707 infections in 2,080 patients — 3.3% of patients, 8.5 per 1,000 patient-days, with a median stay of 12 days. Nearly half were not device-related, and C. difficile had become the most common microbial cause.

   Pombo DJ, Lopansri BK, Keane M, Taylor C, Kleckner R, Sumner S, Mehta RR, Burke JP. IDSA, 2013.

3. 03Infectious disease

   The under-estimated cost of catheter-associated UTI

   1.2% vs 5.6%

   of Medicare encounters coded for the infection, against those who actually developed it

   When Medicare stopped paying for hospital-acquired infections, did the coded record reflect how often they happened?

   Data, method, and finding＋

   Data

   56,554 patients with an indwelling catheter at two tertiary hospitals, 2000–2008, including 28,888 Medicare patients.

   Method

   Matched case–control: 277 coded cases matched on sex, facility, surgery, admission year, severity of illness, and DRG.

   Finding

   382 patients were coded; 1,604 developed the infection. Cases stayed longer (14.2 against 11.0 days) and cost more ($16,913 against $15,553 median per encounter). Cost estimates built on coded cases understate the real burden.

   Dascomb K, Mehta RR, Pombo DJ, Burke JP. IDSA 47th Annual Meeting, 2008.

4. 04Infectious diseaseNull result — published

   Rapid versus overnight bacterial identification

   No difference

   in length of stay, variable cost or mortality

   Would switching to a rapid identification platform deliver the shorter stays and lower costs reported elsewhere?

   Data, method, and finding＋

   Data

   1,959 inpatients at two hospitals — 653 cases and 1,306 matched controls — January 2002 to August 2003.

   Method

   Retrospective case–control, two controls per case matched on sex, APR-DRG severity and specimen type. Kaplan–Meier analysis of length of stay, adjusted for patients who died.

   Finding

   Turnaround was 61 against 62 hours, median stay 7.2 against 7.7 days, and median variable cost $6,875 against $7,566 — none significant. At day 25, 13% against 16% remained in hospital (log-rank 1.68, not significant).

   Mehta RR, Burke JP, Pombo DJ. ICAAC, poster D-697.