Aim: Measure the incidence of iatrogenic complications

Cases: Iatrogenic complications are split into twelve sub-indicators to focus more specifically on possible improvement measures: shock (SH), surgical infections (SI), new born (NB), obstetrics (OB), hemorrhage (HE), thromboembolic (TE), other surgical (SU), skin ulcer (UL), other infections (OI), anesthesia (AN), drugs (DR), health care (HC). All complications have been taken into account. One stay can comprise multiple complications but a main complication (CP) is always defined (the list above is provided in the decreasing order of importance).

Complications are weighted by scores depending on the context in which they occur: premature death (10), potentially avoidable readmission or reoperation (4), length of stay over expected values (2) and others (1). If a complication is only suspected (i.e. not clearly stated as complication during the coding process), scores are halved.

Eligible population: For most of the complications (SH, HE, TE, UL, OI, DR, HC), all stay stays are included in the eligible population. For complication NB and OB, the eligible population is restricted to new born and obstetrical stays (main categories beginning with B-*, F-pA, F-pG, F-pP, F-pV, PAR2 and VAG2). For complications SI, SU and AN, the eligible population is restricted to surgery (main categories listed in procedures).

Expected values: Expected values are computed separately for every sub-indicator since the risks factors depend on the nature of the complications.

Output files: Results by hospital and site are provided for each type of complication (complication output Excel files). Detailed output by hospital stay are provided in the text file Analysis.txt. General information about all SQLape output files can be found here.

Strength of the indicator: Each complication can be attributed to different aspects of health care management such as hospital hygiene, prevention of infections, and surgical practice for instance. Proper adjustment and definition of the population at risk yields precise and unbiased results. Unlike alternative measures of complications, diagnoses associated to complications are excluded from the adjustment model (expected values calculation), which alleviate the issue of over-fitting.

Limitations: Sensitivity of the indicators should be good if complications are coded. High complications rates might reflect an over-coding of such issues, especially for drugs and health care complications. However, weighting complications according to their context (premature death, readmission, reoperation, lengthening the length of stay) enables to reduce the impact of an exaggerated coding of complications, even the most minor ones. Moreover, the measurement of the excess length of stay observed among the cases with a complication enables to evaluate its severity (see indicators on beds).