Advertisement

Direct health care costs and length of hospital stay related to health care-acquired infections in adult patients based on point prevalence measurements

Open AccessPublished:March 14, 2016DOI:https://doi.org/10.1016/j.ajic.2016.01.035

      Highlights

      • All admitted patients from 10 independent Point Prevalence Surveys (PPS).
      • Actual direct healthcare costs were calculated, not templates, nor estimates.
      • The 30-days Follow-up showed more re-admissions in the group with HAI.
      • Patients with HAI had a greater portion of bed days and costs than average.
      • Patients with HAI had a higher one year mortality rate than other patients.

      Background

      The incidence of health care-acquired infection (HAI) and the consequence for patients with HAI tend to vary from study to study. By including all patients, all medical specialties, and performing a follow-up analysis, this study contributes to previous findings in this research field.

      Methods

      Data from the Swedish National Point Prevalence Surveys of HAI 2010-2012 was merged with cost per patient data from the county Health Care Register (N = 6,823). Extended length of stay (LOS) and costs related to an HAI were adjusted for sex, age, intensive care unit use, and surgery.

      Results

      Patients with HAI (n = 732) had a larger proportion of readmissions compared with patients with no HAI (29.0% vs 16.5%). Of the total bed days, 9.3% was considered to be excess days attributed to the group of patients with an HAI. The excess LOS comprised 11.4% of the total costs (95% CI, 10.2-12.7). The 1-year overall mortality rate for patients with HAI in comparison to all other patients was 1.75 (95% CI, 1.45-2.11), all 5 of these differences were statistically significant (P < .001).

      Conclusions

      Even if not all outcomes for patients with an HAI can be explained by the HAI itself, the increase in inpatient days, readmissions, associated costs, and higher mortality rates are quite notable.

      Key Words

      A health care-acquired infection (HAI) is defined as an infection occurring during the process of care in a hospital or other health care facility, neither present nor incubating at the time of admission nor at the time of a visit to a health care facility.
      • Allegranzi B.
      Report on the burden of endemic health care-associated infection worldwide.
      HAIs are known to comprise the largest part of adverse events in health care and cause prolonged hospital length of stay (LOS) and deaths.
      • Allegranzi B.
      Report on the burden of endemic health care-associated infection worldwide.
      • Burke J.P.
      Infection control—a problem for patient safety.
      • Rutberg H.
      • Borgstedt Risberg M.
      • Sjodahl R.
      • Nordqvist P.
      • Valter L.
      • Nilsson L.
      Characterisations of adverse events detected in a university hospital: a 4-year study using the Global Trigger Tool method.
      • Graves N.
      • Weinhold D.
      • Tong E.
      • Birrell F.
      • Doidge S.
      • Ramritu P.
      • et al.
      Effect of healthcare-acquired infection on length of hospital stay and cost.
      In earlier studies, point prevalence surveys have been used to provide information on the presence of HAIs, and costs have been estimated indirectly from the extension of LOS and the mean cost of hospital days in different specialties.
      • Graves N.
      • Weinhold D.
      • Tong E.
      • Birrell F.
      • Doidge S.
      • Ramritu P.
      • et al.
      Effect of healthcare-acquired infection on length of hospital stay and cost.
      • Andersen B.M.
      • Rasch M.
      • Hochlin K.
      • Tollefsen T.
      • Sandvik L.
      Hospital-acquired infections before and after healthcare reorganization in a tertiary university hospital in Norway.
      • Scheel O.
      • Stormark M.
      National prevalence survey on hospital infections in Norway.
      Depending on the type of HAI and the mix of patients included, previous studies have shown different results regarding LOS resulting from HAI. Plowman et al
      • Plowman R.
      • Graves N.
      • Griffin M.A.
      • Roberts J.A.
      • Swan A.V.
      • Cookson B.
      • et al.
      The rate and cost of hospital-acquired infections occurring in patients admitted to selected specialties of a district general hospital in England and the national burden imposed.
      found that HAIs in general extended the LOS by 14 days, which was almost 3 times longer than the average LOS for patients without an HAI. There was a large variation in extended stay depending on the site of single infection (2-13 days) and admission specialty (1-23 days).
      Sheng et al
      • Sheng W.H.
      • Wang J.T.
      • Lu D.C.
      • Chie W.C.
      • Chen Y.C.
      • Chang S.C.
      Comparative impact of hospital-acquired infections on medical costs, length of hospital stay and outcome between community hospitals and medical centres.
      found that the additional LOS was around 20 days for patients affected by an HAI. They also addressed a difficulty regarding estimating costs for HAIs among inpatients by pointing out a higher mortality rate among patients with an HAI than among uninfected patients. Death reduces the direct medical costs but represents for each patient a unique loss of potential life years.
      Klevens et al
      • Klevens R.M.
      • Edwards J.R.
      • Richards Jr, C.L.
      • Horan T.C.
      • Gaynes R.P.
      • Pollock D.A.
      • et al.
      Estimating health care-associated infections and deaths in U.S. hospitals, 2002.
      estimated a US case fatality rate of HAI in hospitalized patients to 5.7% during 1999-2002, with the highest mortality rate in ventilator-associated pneumonia (14.4%) and catheter-associated bloodstream infections (12.3%). In Europe the 30-day mortality rate for Clostridium difficile infection (CDI) is estimated to be from 3%-30% in different countries.
      • Wiegand P.N.
      • Nathwani D.
      • Wilcox M.H.
      • Stephens J.
      • Shelbaya A.
      • Haider S.
      Clinical and economic burden of Clostridium difficile infection in Europe: a systematic review of healthcare-facility-acquired infection.
      Kaye et al
      • Kaye K.S.
      • Marchaim D.
      • Chen T.Y.
      • Baures T.
      • Anderson D.J.
      • Choi Y.
      • et al.
      Effect of nosocomial bloodstream infections on mortality, length of stay, and hospital costs in older adults.
      showed that the 90-day mortality rate in hospitalized older adults rises from 33.2%-49.4% (odds ratio, 2.1; P < .001) in patients with a health care-associated bloodstream infection.
      One challenge in estimating the effect of the additional LOS for patients with an HAI is to extract the effect of the infection from that of any other complications or disease progression. Regression analysis with adjustment for patient characteristics has therefore been a way to clear the effects of an HAI.
      • Graves N.
      • Weinhold D.
      • Tong E.
      • Birrell F.
      • Doidge S.
      • Ramritu P.
      • et al.
      Effect of healthcare-acquired infection on length of hospital stay and cost.
      • Plowman R.
      • Graves N.
      • Griffin M.A.
      • Roberts J.A.
      • Swan A.V.
      • Cookson B.
      • et al.
      The rate and cost of hospital-acquired infections occurring in patients admitted to selected specialties of a district general hospital in England and the national burden imposed.
      Readmissions are more common for patients with an HAI. Chopra et al
      • Chopra T.
      • Neelakanta A.
      • Dombecki C.
      • Awali R.A.
      • Sharma S.
      • Kaye K.
      • et al.
      Burden of Clostridium difficile infection on hospital readmissions and its potential impact under the Hospital Readmission Reduction Program.
      found that it was about twice as common that patients with CDI had a readmission within 30 days compared with all other patients (30.1% vs 14.4%).

       Aim

      The aim of this study was to examine the effects of HAI by calculating the difference in LOS and actual direct health care costs for patients with an HAI compared with patients without HAI. Additional LOS and costs resulting from any first hospital readmission within 30 days from discharge were also included in the analysis, as was the survival rate (as mortality ratio) between the 2 groups 1 year after the primary discharge.

      Materials and methods

       Setting

      The County of Östergötland has the fourth largest population in Sweden, with 434,000 residents and the Region Östergötland (formerly the County Council) both provides and is responsible for all public health care in the county. All patient encounters, including primary care, are registered in the Health Care Register (HCR). The county has 3 hospitals, 1 of which is a university hospital.

       The Point Prevalence Surveys of HAI

      Directed by the Swedish Association of Local Authorities and Regions (SALAR) and using the Swedish National Board of Health and Welfare's definition of HAI, the Point Prevalence Surveys of HAI (PPS-HAI) have been conducted in the majority of public somatic hospitals in Sweden every spring and fall since 2008 and around 20,000 patients are included each semester. The PPS-HAI was conducted with national standard protocols and instructions by SALAR and was performed by nurses and attending physicians on each ward during all somatic clinics. The team that makes the registrations during the days the PPS is administered has to read and understand the instructions (from SALAR) detailing how to perform the survey.
      All inpatients on a somatic ward on a given day were included. Demographic data (age, identity, and sex) were recorded, as was information on medical specialty. HAI was defined as an intention-to-treat diagnosis or final diagnosis of postoperative infection within 30 days from surgery or within 1 year if implantation surgery; device-related infections due to central venous catheters, urinary tract catheters, ventilator treatment, or endotracheal tubes or indwelling cerebral ventricular drainage; drug-related infections defined as C difficile enteritis, infections related to chemotherapy for cancer, or infections due to other immune-modulating drugs or corticosteroids; and other infections occurring > 48 hours after admission. The HAI was categorized according to the suspected anatomic site of the infection origin.

       The Point Prevalence Register

      The Point Prevalence Register (PPR) used includes local data from the County of Östergötland recorded in the national surveys for 2010-2012, together with data from 4 extra surveys conducted in the same way during 2012. The data encompass 10 separate prevalence measurements with an average of 800 patients per registration (range, 729-829). Of all unique patients in this study, 58% were inpatients at the university hospital. The registrations included nonresidents in the county.

       The validation of the national PPS for HAI in hospital care during autumn 2012

      On commission of SALAR, the first author (MR) conducted a national validation study in 2013. The validation study design was to let an independent team consisting of specialists in infectious diseases do the same survey as the ordinary team on the same day. The primary aim was to see how well the results correlated. The study involved 1,216 patients across 20 hospitals in Sweden, the specialties chosen for the study were general surgery and internal medicine. In total, the independent teams found more patients with an HAI than the ordinary teams did (8.3% in the ordinary survey, whereas the validation teams found a prevalence of 13.1%) (95% confidence interval [CI], 11.2-15.0) (Table B in the Supplementary material). All registered cases of HAI became 15.3% when the figures from the 2 independent registrations were added together (Table A in the Supplementary material).
      On the local level the validating teams found a higher prevalence at 17 hospitals, the same prevalence at 2 hospitals, and a lower level at 1 hospital. The conclusion from the validation study was that the ordinary teams in most cases underestimated the true prevalence. Details of the validation study are provided as supplementary material.

       The HCR

      Record data on medical specialty, main diagnosis, LOS, costs, and other factors for each inpatient in somatic care were retrieved from the HCR. All visit or hospitalization records include a personal identification number (PID), date of visit or hospitalization, day of discharge, and diagnosis according to ICD-10,
      • World Health Organization
      ICD-10 International Statistical Classification of Diseases and Related Health Problems 10th Revision.
      with main diagnosis and secondary diagnosis. The patient-specific medical costs for visits and inpatient care in the HCR were priced with the additional module of cost per patient containing the diagnosis-related group costs per patient, as described in other studies.
      • Wirehn A.B.
      • Andersson A.
      • Ostgren C.J.
      • Carstensen J.
      Age-specific direct healthcare costs attributable to diabetes in a Swedish population: a register-based analysis.
      The detailed cost figures in the cost per patient module (ie, pricing) are administrated by SALAR.
      There were 8,104 registrations with a valid PID in the PPR, but 59 of them were double registrations on the same PPS for patients with >1 HAI, thus leaving 8,045 unique registrations. However, 27 of these cases were doubled for patients who had been registered on >1 PPS during the same admission, leaving 8,018 unique patients to be matched with the HCR. When we merged PPR data with HCR data, 37 patients from the PPR were impossible to find in the HCR. For these cases, we suspect that an incorrect PID had been entered in the PPR, and we found no systematic selection bias for the 23 patients for whom we could find no information. These 23 (of 37) patients were completely random across surveys and medical departments.
      The complete and valid records resulted in 7,981 patients, of whom 316 died during admission (Table 1). In the 30-day follow-up of readmissions and in the calculation of 1-year survival after the discharge of the PPS admission the study population was reduced to 7,062 patients, of whom 6,823 were adults. Only residents of the county who were registered in the HCR could be part of the follow-up, and only patients who survived the original admission and the 30-day follow-up could be included in the final analysis.
      Table 1Number of patients in the dataset and test for health care-acquired infection (HAI) prevalence differences, by subgroup
      GroupnHAI, nHAI, %P value for HAI vs totalMean LOSMedian LOSMean cost €1,000
      Unique registrations in PPR (1 PID/admission) found in the HCR7,98186010.812.0715.1
       Men all ages3,85848112.8.00612.8717.9
       Women all ages4,1233799.2−.00311.2712.5
       Patients who died during the hospitalization (excluded)−3165517.4< .00120.61528.3
       Patients who died within 30 d after discharge (excluded)−3415415.8.00317.41316.0
       Patients with >90 days of hospitalization (excluded)−462452.2< .001136.9113384.6
       Nonresidents all ages discharged alive (excluded)−2213415.4.03712.4725.3
      Remaining patients for main analyses
       Patients/residents in the follow-up with an LOS during PPS ≤ 90 d7,0626979.9ns10.6611.8
       Children/residents aged 0-17 y alive >30 d from discharge23983.3−< .0016.449.8
       Adult patients/residents alive >30 d from discharge6,82368910.1ns10.7711.9
      Readmissions
       All readmissions within 30 d from discharge1,40319814.1< .00113.6815.7
       Valid readmissions within 30 d from discharge (alive and resident)1,25717714.1< .00113.4815.9
      HCR, Health Care Registry; LOS, length of stay; ns, not significant; PID, personal identification number; PPS, point prevalence survey; PPR, Point Prevalence Register.

       Direct cost calculations

      The costs are expressed in Euros, in the price level of 2012. The costs for 2010 and 2011 were inflated with the consumer price index to reflect 2012 prices. The exchange rate from Swedish crowns (SEK) to Euros was set to SEK 9 per Euro. To convert the costs to US dollars, use a factor of 1.3 times the cost in Euros, derived from the monthly averages of exchange SEK to US dollars during 2012, a year average of SEK 6.8 per US dollar.

       Study design

      The flow chart in Figure 1 illustrates how each patient can have a unique date for admission, but share the PPS date with other patients for that particular survey. Ten independent surveys were performed during the study period 2010-2012. Of 7,062 eligible patients 1,245 had 1 or more readmissions within 30 days from discharge from the original admission. Only any first readmission within 30 days from discharge was included in the calculations of total LOS and costs.
      Figure thumbnail ymic3860-fig-0001
      Fig 1The study design illustrated in a flow chart for admissions, how health care-acquired infections (HAIs) are registered, and when survival status is checked. Survival status is checked 365 days from the discharge date of the first and original admission. LOS, length of stay; PPS, Point Prevalence Survey.
      The prevalence rate of HAI in Table 2 was created by adding any new cases of HAI that were registered with ICD-10 codes during the readmission (n = 50), yielding a total of 740 patients with HAI of whom 732 were adults. The mortality rate was calculated for a period of 365 days, starting on the date of discharge from the original admission. However, patients who died during the admission or the follow-up period were excluded from the regression analyses and the mortality rate calculations. Therefore, all patients in the regression were alive 30 days after the discharge from the original admission.
      Table 2Total bed days and percent excess days related to patients with a health care-acquired infection (HAI)
      nHAI, nHAI, %Number of daysPercent of days for the group with HAIAdjusted portion of excess days for the group with HAI,
      Excess costs adjusted for sex, age, intensive care unit use, and surgery.
      %
      Total

      For all
      Group HAITotalUnadjusted excess HAIAdjusted %95% Confidence intervalP value
      All patients of all ages with a Point Prevalence Survey admission7,06274010.574,65714,51719.410.09.110.0-12.9<.001
      All patients all ages with any first readmission7,06274010.583,83816,76220.010.69.411.1-13.9<.001
      Children/residents aged 0-17 y23983.31,69218811.18.07.02.8-14.6.004
      Adult patients aged 18 y and older
       All adult patients6,82373210.782,14616,57420.210.69.38.4-10.2<.001
       Ear, nose, and throat11697.888121624.518.217.511.4-23.5<.001
       Gynecology498224.42,60029211.27.16.23.6-8.6<.001
       Infection2835820.54,9301,35627.58.87.92.5-13.4.005
       Internal medicine3,3443249.745,9047,76416.98.07.66.5-8.8<.001
       Orthopedics906879.68,5231,98623.311.912.310.0-14.5<.001
       Surgery1,41418913.417,0314,50226.415.112.310.3-14.5<.001
       Urology2624316.42,54772828.614.513.26.9-19.6<.001
       Patients without intensive care unit stay6,3156179.871,23812,84518.07.08.87.9-9.8<.001
       Patients with intensive care unit stay50811522.610,9073,72734.212.213.59.8-17.3<.001
       Patients without surgery6,0465629.365,70111,35517.37.48.87.8-9.6<.001
       Patients with surgery77717021.916,4455,19831.610.811.78.6-14.7<.001
      * Excess costs adjusted for sex, age, intensive care unit use, and surgery.

       Statistical analysis

      The χ2 test was used for comparing the prevalence of HAI and gender differences of subgroups versus the total, and also for mortality ratios of deaths during admission. Student t test was used for average age comparison, whereas linear regression analysis was used for the main outcome of the adjusted proportions for excess bed days and direct medical costs. Any differences in 1-year survival were tested with multinomial regression on mortality ratio, together with standardized mortality ratio (SMR) based on the expected outcome for the reference population who did not experience HAI. P values < .050 were considered statistically significant. All statistics and dataset arrangements were done with IBM SPSS version 22 and version 23 (IBM-SPSS Inc, Armonk, NY).

       Regression analysis including 30-day readmission follow-up

      In the follow-up analysis, LOS and cost per patient in all patients residing in the county eventually readmitted within 30 days from discharge and alive during the follow-up period (n = 7,062), were included. Nonresidents (n = 226) were excluded because of a lack of data regarding continued hospital care and readmissions in their home county. Only the first readmission was included because several patients had >1 readmission within 30 days. Data for the adult patients were stratified by medical specialty (Table 2 and Table 3).
      Table 3Total costs and percent excess costs related to patients with a health care-acquired infection (HAI)
      No HAI, nHAI, nHAI, %Costs in €1,000Percent of costs for the group with HAIAdjusted portion of excess costs for the group with HAI,
      Excess costs adjusted for sex, age, intensive care unit use, and surgery.
      %
      Total for allGroup HAITotalUnadjusted excess HAIAdjusted %95% Confidence intervalP value
      All patients of all ages with a Point Prevalence Survey admission6,32274010.583,57018,64822.313.210.38.9-11.5<.001
      All patients of all ages with any first readmission6,32274010.595,97921,76022.713.610.99.7-12.2<.001
      Children/residents aged 0-17 y23183.32,8052498.95.75.21.7-8.8.004
      Adult patients aged 18 y and older0
       All adult patients6,09173210.793,17521,51123.113.811.410.2-12.7<.001
       Ear, nose, and throat10797.81,08317816.89.58.71.7-15.9.016
       Gynecology476224.42,54938014.911.09.87.5-12.0<.001
       Infection2255820.54,8231,57032.615.213.76.2-21.1<.001
       Internal medicine3,0203249.744,9848,48218.910.29.57.7-11.2<.001
       Orthopedics819879.68,5191,43016.88.07.94.6-11.3<.001
       Surgery1,22518913.428,4998,69030.519.815.212.7-17.9<.001
       Urology2194316.42,71477628.614.613.75.2-22.2.002
       Patients without intensive care unit stay5,6986179.868,89913,02218.910.19.58.2-10.6<.001
       Patients with intensive care unit stay39311522.624,2758,48834.715.926.518.5-34.5<.001
       Patients without surgery5,4845629.361,67010,81817.59.19.07.8-10.3<.001
       Patients with surgery60717021.931,50410,69333.915.424.017.8-30.2<.001
      * Excess costs adjusted for sex, age, intensive care unit use, and surgery.

       Group of HAI and the proportion of excess bed days

      The 4 main types of HAI registered in the PPS were postsurgery infection, other invasive procedure-related infections, pharmaceutical-related infections, and other infections (Table 4). The proportion of excess bed days for these groups were analyzed in the same manner as across specialty.
      Table 4Total bed days and percent excess days related to patients with a health care-acquired infection (HAI) across type of HAI
      Adult patients aged 18 y and oldernHAI, nHAI, %Number of bed daysPercent of days for the group with HAIAdjusted portion of excess days for the group with HAI,
      Excess costs adjusted for sex, age, intensive care unit use, and surgery.
      %
      TotalGroup HAITotalUnadjusted excess HAIAdj. %95% Confidence intervalP value
      Postsurgery infection6,8232253.382,1464,9176.03.12.42.1-3.0< .001
      Other invasive procedure-related infections6,8231852.782,1463,9694.82.42.01.5-2.4< .001
      Pharmaceutical-related infection6,8231582.382,1463,9084.82.72.72.3-3.1< .001
      Other infections6,8231642.482,1463,7804.62.52.31.9-2.6< .001
      Total6,82373210.682,14616,57420.210.79.38.4-10.2< .001
      * Excess costs adjusted for sex, age, intensive care unit use, and surgery.

       The 1-year mortality rate test

      The HCR is linked to the vital statistics for the county population and for the patients who died within 1 year from the original admission discharge the date information was entered to the mortality rate analysis. The SMR is presented together with 2 nominal regressions of the 1-year survival (mortality ratio), first with age and sex as adjusting factors, followed by a second analysis where any presence of intensive care unit stay or surgery also were added into the regression as adjusting factors for survival.

       Ethics approval

      This study was approved by the Regional Ethics Committee at Linköping University, Sweden (2013/196-31).

      Results

      The average prevalence of HAI for all patients (N = 7,981) was 10.8%. For patients in the main analyses (n = 7,062) the prevalence of HAI in the PPS was 9.9%. For this group of patients and residents there was a variation of prevalence of HAI in the 10 independent PPSs between 8.7% and 13.1%, with an average of 9.9%. Of the total patient population in somatic care there were fewer men compared with women (48.3% vs 51.7%; P = .008). However, the proportion of men was greater in the following groups (P < .01): patients with >90 days hospitalization (76%), children alive >30 days from discharge (58%), and in the group that had at least 1 readmission within 30 days from discharge (53%).
      Men had a significantly higher prevalence of HAI than women (Table 1). Patients who died during the hospitalization or within the 30-day follow-up had a higher prevalence of HAI than average (17.4% and 15.8%, respectively). Children had a lower prevalence of HAI (4.3%), whereas nonresidents had a higher prevalence of HAI than average (15.2%) (Table 1). During the admission 6.4% of patients with an HAI died compared with 4.4% of the other patients, which was a statistically significant difference according to Mantel-Haenzel test (χ2 = 6.6; P = .005).
      Of 7,062 patients/residents alive during the 30-day follow-up period, 1,245 had a readmission within 30 days. Patients with HAI (n = 732) had a larger proportion of readmissions compared with patients with no HAI (29.0% vs 16.5%). Half of all readmissions (49.6%) were admitted within 10 days of discharge. Of the readmitted patients, 43 new patients received a diagnosis that explicitly indicated they experienced an HAI that was not registered in the PPS. Within the top-10 reasons for a readmission, we noted the following 4 ICD-10 groups, each group to a varying extent related to a previous care episode: T81.4 “Infection following a procedure, not elsewhere classified”; T81.8 “Other complications of procedures, not elsewhere classified”; A04.7 “Enterocolitis due to Clostridium difficile”; and J18.9 “Pneumonia, unspecified.” The top-10 diagnoses accounted for 17.9% of all readmitted cases. It was almost twice as common that patients with a registered HAI had a readmission (29.2%) compared with other patients (16.3%) (n = 216 and n = 1,029, respectively).
      The results in adjusted LOS and excessive costs for adult patients are presented by specialty and subgroup as percentage of the total bed days and costs. Adjustment for age, sex, intensive care unit use, and surgery was done for several reasons: patients with an HAI were older than patients without an HAI (P = .001), men had a significantly higher prevalence of HAI (P <.001), the prevalence of HAI was significantly higher among patients who had received intensive care, and the group with surgery had a statistically significantly higher prevalence of HAI than the group without surgery (P <.001) (Table 1). For adult patients (n = 6,823) (see Table 2 and Table 3) a total of 7.4% had a period of intensive care unit use, and a total of 11.4% underwent surgery.
      In the follow-up analysis including readmission (n = 6,832), the prevalence of HAI among adult patients was 10.7% (n = 732). The total number of bed days for adult patients without an HAI comprised 79.8% of all bed days, and patients with an HAI accounted for 20.2% of all days. Of all days, 9.3% were considered to be excess days for the group with an HAI (Table 2) and the adjusted average of excess days were 10.5 per patient with an HAI (95% CI, 9.5-11.5). The cost for the excess days was, after adjustment, 11.4% of the total costs (95% CI, 10.2-12.7) (Table 3). The difference across medical specialties was large, ranging from 6.2% of the subtotal days (infection speciality) to 17.5% of the subtotal days (ear, nose, and throat specialty) (Table 2). The share of excessive costs compared with the subtotal costs for patients with HAI varied between 7.9% (orthopedics specialty) and 26.5% (surgery specialty) (Table 3).
      The analyses of the 4 main types of HAI show that patients with infections related to pharmaceutical treatment were the group that added most to the proportion of adjusted excess bed days (2.7%), despite the fact that this group was the smallest of the 4 groups (Table 4).
      The 1-year survival was calculated from the day of discharge from the first admission (the admission when the PPS was completed) for patients who were alive during the admission and the 30-day follow-up. One year from discharge, 15.2% of patients who had no HAI had died compared with 24.5% of patients with an HAI. This gave an SMR of 1.61 (95% CI, 1.39-1.86; P < .001) (Fig 2). The nominal regression for 1-year survival gave an adjusted mortality ratio of 1.75 for patients with HAI compared with other patients, which was statistically significant with a P value <.001 (95% CI, 1.45-2.11). This regression was adjusted for age and sex, but in the extended nominal regression intensive care unit stay and surgery was also adjusted for. Including intensive care and surgery into the regression slightly increased the mortality ratio for patients with an HAI to 1.81 (95% CI, 1.50-2.19).
      Figure thumbnail ymic3860-fig-0002
      Fig 2One-year survival rate from the primary discharge (%) for adult patients who survived the 30-day follow-up period. Patients who had a primary length of stay of >90 days are excluded. HAI, health care-acquired infection.

      Discussion

      Adult patients registered with an HAI encompassed about a 10th of all admitted patients in our study. However, the PPS-HAI figure is not the true prevalence of HAI, but shows the proportion of hospital beds occupied by patients with an HAI during a given day. To include all patients in a hospital during 1 day, as done in a PPS, favors patients with long hospital stays, whereas a representative sample would not. Despite this length bias we consider the patients with or without a HAI in this study as 2 comparable groups. Thereby any differences in the outcomes shall be regarded as true differences, but LOS and costs are not representative of the whole patient population. The criteria for inclusion in the analyses (ie, excluding children, patients who had a LOS > 90 days, and patients who died before the 30-day follow-up period ended) also contributes to reducing the generalizability to a complete patient population. Some of these exclusions reduce the influence from outliers who have extraordinary long treatment periods and associated extreme costs. And by concentrating on adult patients, we excluded outliers with more than 3 months' LOS, and also only include patients who survived the admission and the following month, we believe that we to some degree deflate the overall effects of HAI, and also increase the generalizability to other studies that will use the same design.
      The use of intention-to-treat diagnoses supplemented with ICD-10 diagnoses indicating HAI probably increases the incidence of HAI, unlike the use of well-defined final diagnoses,
      • Horan T.C.
      • Andrus M.
      • Dudeck M.A.
      CDC/NHSN surveillance definition of health care–associated infection and criteria for specific types of infections in the acute care setting.
      although the use of final diagnoses could be criticized for underestimating the number of patients with HAI and thereby the costs.
      In the follow-up analysis with any first readmission, we found that the total number of bed days increased with 14.2% for the group with a HAI compared with 11.5% for the group without any HAI. To include readmission is a rather new and a most-relevant approach in this research field, because an HAI may first become observable after discharge. By including readmissions, a more complete picture of patient-related LOS and costs can be achieved. Just like Chopra et al
      • Chopra T.
      • Neelakanta A.
      • Dombecki C.
      • Awali R.A.
      • Sharma S.
      • Kaye K.
      • et al.
      Burden of Clostridium difficile infection on hospital readmissions and its potential impact under the Hospital Readmission Reduction Program.
      in their study regarding CDI, we found that it was about twice as common that patients with an HAI had a readmission within 30 days compared with all other patients, which confirms the relevance of adding the LOS and costs for any first readmission (or several readmissions) when studying the effects of HAI.
      The significant increase of inpatient days following an HAI that we found is in line with previous studies
      • Plowman R.
      • Graves N.
      • Griffin M.A.
      • Roberts J.A.
      • Swan A.V.
      • Cookson B.
      • et al.
      The rate and cost of hospital-acquired infections occurring in patients admitted to selected specialties of a district general hospital in England and the national burden imposed.
      • Sheng W.H.
      • Wang J.T.
      • Lu D.C.
      • Chie W.C.
      • Chen Y.C.
      • Chang S.C.
      Comparative impact of hospital-acquired infections on medical costs, length of hospital stay and outcome between community hospitals and medical centres.
      • Soop M.
      • Fryksmark U.
      • Koster M.
      • Haglund B.
      The incidence of adverse events in Swedish hospitals: a retrospective medical record review study.
      ; however, we avoided counting days per patient in favor of group statistics. One great challenge when estimating the effect of LOS for patients with an HAI is to extract the effect of the infection and not exaggerate the LOS from any other complications or disease progression. A straight-matched, case-control study can reduce the effects of confounding factors, and DiGiovine et al
      • Digiovine B.
      • Chenoweth C.
      • Watts C.
      • Higgins M.
      The attributable mortality and costs of primary nosocomial bloodstream infections in the intensive care unit.
      present a good example where severity of illness was accounted for, and they found that the excess LOS was 10 days for patients with primary nosocomial bloodstream infections. The approach of Vrijens et al
      • Vrijens F.
      • Hulstaert F.
      • Devriese S.
      • van de Sande S.
      Hospital-acquired infections in Belgian acute-care hospitals: an estimation of their global impact on mortality, length of stay and healthcare costs.
      to the problem was both to concentrate on 1 type of HAI—bloodstream infection—and to study the outcome on a matched cohort. The adjusted mean difference in LOS between infected and uninfected patients was reduced by about 20% when matched for primary diagnosis and comorbidity. We had no such sophisticated data, but made adjustments for age, sex, intensive care unit use, and surgical operations, which significantly reduced the average difference on the total level. To include more background factors and risk factors would reduce the differences between the groups, but such an inclusion would on the other hand make the interpretation of the outcomes more difficult.
      Comorbidity might contribute to a longer hospital stay for a patient with an HAI (above average for patients with an HAI with no comorbidity) but if the patient with comorbidity was not affected by an HAI in the first place, then the hospital stay would have been shorter, at best the average length for patients without any HAI. The differences between infected and uninfected patients in this study do not show a pure effect of the HAI. Although, as could be seen in the validation study, prevalence rates are often underestimated in the majority of the Swedish PPS for HAI, and if there were more infected patients in our data, the percentage of days and costs would increase.
      Including figures for readmission and overall mortality up to 12 months from discharge gives this study a rather complete and unique picture of the patients registered in the initial prevalence surveys. Most studies that report mortality, more than in-hospital mortality, use a 30-day period and of 56 studies that reported mortality for patients with CDI, only 4 studies followed their patients for 12 months,
      • Wiegand P.N.
      • Nathwani D.
      • Wilcox M.H.
      • Stephens J.
      • Shelbaya A.
      • Haider S.
      Clinical and economic burden of Clostridium difficile infection in Europe: a systematic review of healthcare-facility-acquired infection.
      indicating that 12-month mortality figures are rare because in the same systematic review 14 studies had figures only for 30-day mortality, whereas the 38 remaining studies had figures for in-hospital mortality and/or mortality up to 60 or 90 days. According to the same influence that risk factors and comorbidity have on the outcomes of reduced number of days and costs, the difference in mortality rate presented in this study would decrease significantly if patient risk factors was entered into the survival regression. Unfortunately we lack information concerning relevant factors such as comorbidity, health status, body weight, and lifestyles.
      Considering that Harbarth et al
      • Harbarth S.
      • Sax H.
      • Gastmeier P.
      The preventable proportion of nosocomial infections: an overview of published reports.
      • Lambert M.L.
      • Silversmit G.
      • Savey A.
      • Palomar M.
      • Hiesmayr M.
      • Agodi A.
      • et al.
      Preventable proportion of severe infections acquired in intensive care units: case-mix adjusted estimations from patient-based surveillance data.
      estimated the percentage of all preventable HAIs at about 15%, and that Lambert et al
      • Harbarth S.
      • Sax H.
      • Gastmeier P.
      The preventable proportion of nosocomial infections: an overview of published reports.
      • Lambert M.L.
      • Silversmit G.
      • Savey A.
      • Palomar M.
      • Hiesmayr M.
      • Agodi A.
      • et al.
      Preventable proportion of severe infections acquired in intensive care units: case-mix adjusted estimations from patient-based surveillance data.
      suggested that for specific HAIs such as ventilator-associated pneumonia and bloodstream infections in ICUs 52% and 69%, respectively, are preventable, there is often a potential to reduce the occurrence of HAI. In that light, our results imply that a reduction of HAI prevalence to a significant degree could reduce health care costs, lessen patient suffering, and also increase patients' long-term survival.

      Acknowledgment

      The authors thank L. Svensson and R. Wiklund, Business Intelligence Group, Region Östergötland, for retrieving the HCR and cost per patient records for this study.

      Appendix. Supplementary material

      The following is the supplementary data to this article:

      References

        • Allegranzi B.
        Report on the burden of endemic health care-associated infection worldwide.
        (Geneva Switzerland: World Health Organization)2011
        • Burke J.P.
        Infection control—a problem for patient safety.
        N Engl J Med. 2003; 348: 651-656
        • Rutberg H.
        • Borgstedt Risberg M.
        • Sjodahl R.
        • Nordqvist P.
        • Valter L.
        • Nilsson L.
        Characterisations of adverse events detected in a university hospital: a 4-year study using the Global Trigger Tool method.
        BMJ Open. 2014; 4: e004879
        • Graves N.
        • Weinhold D.
        • Tong E.
        • Birrell F.
        • Doidge S.
        • Ramritu P.
        • et al.
        Effect of healthcare-acquired infection on length of hospital stay and cost.
        Infect Control Hosp Epidemiol. 2007; 28: 280-292
        • Andersen B.M.
        • Rasch M.
        • Hochlin K.
        • Tollefsen T.
        • Sandvik L.
        Hospital-acquired infections before and after healthcare reorganization in a tertiary university hospital in Norway.
        J Public Health (Oxf). 2009; 31: 98-104
        • Scheel O.
        • Stormark M.
        National prevalence survey on hospital infections in Norway.
        J Hosp Infect. 1999; 41: 331-335
        • Plowman R.
        • Graves N.
        • Griffin M.A.
        • Roberts J.A.
        • Swan A.V.
        • Cookson B.
        • et al.
        The rate and cost of hospital-acquired infections occurring in patients admitted to selected specialties of a district general hospital in England and the national burden imposed.
        J Hosp Infect. 2001; 47: 198-209
        • Sheng W.H.
        • Wang J.T.
        • Lu D.C.
        • Chie W.C.
        • Chen Y.C.
        • Chang S.C.
        Comparative impact of hospital-acquired infections on medical costs, length of hospital stay and outcome between community hospitals and medical centres.
        J Hosp Infect. 2005; 59: 205-214
        • Klevens R.M.
        • Edwards J.R.
        • Richards Jr, C.L.
        • Horan T.C.
        • Gaynes R.P.
        • Pollock D.A.
        • et al.
        Estimating health care-associated infections and deaths in U.S. hospitals, 2002.
        Public Health Rep. 2007; 122: 160-166
        • Wiegand P.N.
        • Nathwani D.
        • Wilcox M.H.
        • Stephens J.
        • Shelbaya A.
        • Haider S.
        Clinical and economic burden of Clostridium difficile infection in Europe: a systematic review of healthcare-facility-acquired infection.
        J Hosp Infect. 2012; 81: 1-14
        • Kaye K.S.
        • Marchaim D.
        • Chen T.Y.
        • Baures T.
        • Anderson D.J.
        • Choi Y.
        • et al.
        Effect of nosocomial bloodstream infections on mortality, length of stay, and hospital costs in older adults.
        JAGS. 2014; 62: 306-311
        • Chopra T.
        • Neelakanta A.
        • Dombecki C.
        • Awali R.A.
        • Sharma S.
        • Kaye K.
        • et al.
        Burden of Clostridium difficile infection on hospital readmissions and its potential impact under the Hospital Readmission Reduction Program.
        Am J Infect Control. 2015; 43: 314-317
        • World Health Organization
        ICD-10 International Statistical Classification of Diseases and Related Health Problems 10th Revision.
        (Geneva. World Health Organization; Available from:) (Accessed March 23, 2016)
        • Wirehn A.B.
        • Andersson A.
        • Ostgren C.J.
        • Carstensen J.
        Age-specific direct healthcare costs attributable to diabetes in a Swedish population: a register-based analysis.
        Diabet Med. 2008; 25: 732-737
        • Horan T.C.
        • Andrus M.
        • Dudeck M.A.
        CDC/NHSN surveillance definition of health care–associated infection and criteria for specific types of infections in the acute care setting.
        Am J Infect Control. 2008; 36: 309-332
        • Soop M.
        • Fryksmark U.
        • Koster M.
        • Haglund B.
        The incidence of adverse events in Swedish hospitals: a retrospective medical record review study.
        Int J Qual Health Care. 2009; 21: 285-291
        • Digiovine B.
        • Chenoweth C.
        • Watts C.
        • Higgins M.
        The attributable mortality and costs of primary nosocomial bloodstream infections in the intensive care unit.
        Am J Respir Crit Care Med. 1999; 160: 976-981
        • Vrijens F.
        • Hulstaert F.
        • Devriese S.
        • van de Sande S.
        Hospital-acquired infections in Belgian acute-care hospitals: an estimation of their global impact on mortality, length of stay and healthcare costs.
        Epidemiol Infect. 2012; 140: 126-136
        • Harbarth S.
        • Sax H.
        • Gastmeier P.
        The preventable proportion of nosocomial infections: an overview of published reports.
        J Hosp Infect. 2003; 54 (quiz 321): 258-266
        • Lambert M.L.
        • Silversmit G.
        • Savey A.
        • Palomar M.
        • Hiesmayr M.
        • Agodi A.
        • et al.
        Preventable proportion of severe infections acquired in intensive care units: case-mix adjusted estimations from patient-based surveillance data.
        Infect Control Hosp Epidemiol. 2014; 35: 494-501