SSI recognized after discharge was confirmed in 89 (1.9%) of 4,571 procedures. One hundred seventy-eight patients with similar age, procedure types, and surgical duration were matched to the SSI patients in a ratio of one case-patient to two controls (). No significant differences in age, gender, or surgery type between case-patients and matched controls were noted. Surgery duration was significantly longer for SSI patients, despite having been matched for procedure duration. This was expected because procedure duration is an important risk factor for infection.
Descriptive characteristics of cohort in study of surgical site infections (SSI), Harvard Pilgrim Health Care, 1997–1998a
Impact on Health, Activities, and Perceived Care Needs
One hundred seventy-three (65%) of 267 questionnaires were returned. Those who completed the questionnaire (responders) were slightly older than those that did not respond (58.2 years vs. 54.6 years, p=0.05). No other differences between questionnaire responders and nonresponders were significant ( and ). Among patients who completed the questionnaire, no differences between case-patients and controls were significant for age, sex, and procedure types (), or in the baseline SF-12 assessment of mental and physical health (). Reported occupations of patients and controls did not differ, and few differences between case-patients and controls existed with respect to self-declared differences in pre-existing medical conditions (). Case-patients did experience longer duration of surgery than did controls. Case-patients were also more likely than controls to report a history of congestive heart failure (12% vs. 2.5%, p=0.02) and arthritis (39% vs. 22%, p=0.03). There was a trend towards more case-patients having diabetes than controls (24% vs. 12% p=0.06).
Comparison of questionnaire responders to nonresponders, surgical site infection (SSI) studya
Univariate analysis of questionnaire respondents, surgical site infections (SSIs) studya
In assessing time and productivity costs, we found that case-patients (64%) were more likely than controls (42%) to have spent at least 1/2 day in bed, thus missing planned regular activities (p=0.04). However, differences between case-patients and controls in other areas of lost productivity, such as missed days of work and inability to complete regular activities, were not significant.
Case-patients with an SSI (69%) were more likely than controls (48%) to require home health provider visits (p=0.01). Similar results were found after controlling for age, procedure duration, and baseline SF-12 physical function. There were trends for patients with SSI wanting more home health visits than were provided and wanting a 24-hour hotline to contact a health-care practitioner. Patients, but not controls, reported significantly lower physical health and mental health component scores on the SF-12 after surgery, compared to their own baselines (p=0.003 and p=0.02, respectively).
Health Resource Use in 8 Weeks after Surgery
Patients with SSI recognized after discharge required significantly more resources within the outpatient setting than those without SSI (). Significantly more patients with SSI had at least one ambulatory-care visit, and their average number of visits (7.5) was more than twice the average of those without SSI (3.4). Additionally, case-patients were significantly more likely to call their provider and to make more phone calls to their provider than controls. The number of laboratory tests ordered did not differ between cases and controls. Estimated ambulatory outpatient visits costs generated were on average $365 per case with an SSI and $160 per control during the 8-week postoperative period (p<0.001).
Univariate analysis of 8-week postoperative resource utilization, surgical site infections (SSIs) studya
Patients with an SSI recognized after discharge also used significantly more resources outside of the ambulatory-care centers. More case-patients (31%) had at least one visit to an emergency room compared to controls (9%), p<0.001, and they generated significantly more emergency room charges ($333 vs. $114, p<0.001).
Those with SSI were more likely to require a radiology test (40% vs. 28%, p=0.02) and had higher radiology test charges ($1,076 vs. $587, p=0.02) than those without SSI. More patients with an SSI received durable medical equipment than did controls (37% vs. 22%, p=0.008) and generated higher average durable medical equipment–related charges ($123 vs. $69, p=0.01). A greater proportion of case-patients (62%) than controls (47%) required home health services (p=0.009). Charges related to home health services were higher for those with an SSI ($827) than for those without an SSI ($579), p=0.007. Twice as many case-patients required a stay in a skilled nursing facility (9% vs. 4.5%, p=0.09). There was a nonsignificant trend towards higher average skilled nursing charges for case-patients ($460 vs. $204 p=0.14); however, the average number of days in a skilled nursing facility was the same for case-patients and controls.
Patients with an SSI recognized after discharge generated higher standard wholesale costs for antibiotics than did controls without an SSI. Case-patients had an average cost of $60 for antibiotics, while controls had costs of $13.60 per person (p<0.001). Patients with an SSI were more likely to be readmitted to the hospital (34%) than those without an SSI (12%), p<0.001. These rehospitalizations led to $7,925 charges per person with an SSI compared with charges of $2,079 for those without an SSI (p<0.001). After the conversion of charges to costs, an SSI diagnosed after discharge was associated with excess costs of $2,573 ($3,489 minus $916) from rehospitalization across the entire population who developed an SSI, regardless of readmission status.
Total estimated costs per person incurred during the 8 weeks after discharge from the hospital associated with the index procedures were $5,155 for case-patients with SSI and $1,773 for controls without an SSI (p<0.001). Therefore, costs were $3,382 or 2.9 times greater in patients with SSI recognized after discharge. The subsets of these costs that occurred in those 216 patients never readmitted to any hospital (including the index hospital) were, on average, $928 in case-patients and $621 in controls (p<0.001). Therefore, patients with SSI had on average $307 additional costs that would not have been captured by an infection control surveillance system limited to the inpatient setting. Additionally, in this particular cohort of patients, 23% of all re-admissions and 18% of all emergency room visits occurred at institutions other than the index hospital; such visits and admissions would not have been captured by standard inpatient infection control surveillance.
The mean chronic disease score was significantly higher among case-patients (3,058) than controls (2,148) (p=0.005), as expected on the basis of the higher prevalence of selected chronic diseases in those at risk for an SSI. To determine if preexisting conditions could account for some of the costs associated with SSI recognized after discharge, we used a matched linear regression model; the calculated chronic disease score was the predictor for log-transformed total costs (). Although the chronic disease score was a strong independent predictor of postoperative resource use, even in this matched cohort, it was not a meaningful confounder of the impact of SSI on resource utilization. The parameter estimate for being a case was 1.30 for log-transformed costs in the unadjusted model and 1.20 for log-transformed costs in the adjusted model when chronic disease score was included. This finding suggests that, even after preexisting conditions are adjusted for, SSIs recognized after hospital discharge are significantly associated with higher total costs.
Results of five separate matched linear regression models with log-transformed total costs as the outcome variable, surgical sites infection (SSI) study
Even though we matched case-patients and controls on duration of index surgery, patients with SSI recognized after hospital discharge had significantly longer duration of surgery. To measure if duration of index surgery could confound the total attributable costs of SSI recognized after hospital discharge, we used a matched linear regression model with duration of index surgery and SSI as predictors for log-transformed total costs. The addition of duration of index surgery into the model did not significantly confound the attributable impact that SSI had on higher total costs ().