´óÏóÊÓƵ

In total, we screened nearly 4,000 articles for possible inclusion in meta-analysis—the majority of which were screened out based on a title and abstract review (3,038 of 3,979, or 76.4 percent, were eliminated). Full text review eliminated 740 of the remaining 941 articles (78.6 percent). After further exclusions during data extraction based on the usability of study estimates in meta-analysis, we obtained our final list of articles included in meta-analysis for attributable cost and excess mortality associated with each HAC (Appendix C). In our final estimates, the number of studies included vary between HACs. CDI—the newest addition to the list of HAC—had the most robust body of literature with 9 studies for additional cost and 13 for excess mortality. In contrast, the literature contained far fewer usable studies for falls (3 for costs and 1 for mortality) and OBAE (2 for costs and none for mortality). 

In Exhibit 7, we provide estimates for the additional costs associated with each HAC. The table shows the number of studies included, the range of costs estimates in those studies, and finally, the pooled meta-analysis based estimate of additional costs with a 95% confidence interval. The 95% CI arises from a two-sided test of the hypothesis that the estimate of additional costs does not differ from a value of zero dollars. When this CI does not include zero, we can assume the HAC does have additional costs associated with its treatment above and beyond the costs for a hospital stay for similar patients without the HAC.

We define additional cost as the incremental costs to the hospital for the inpatient stay attributable with the HAC of interest. The costs are limited to the hospital costs that would not have occurred had the HAC not occurred. These estimates do not include related costs (e.g., days of lost work) or costs of a readmission resulting from the HAC. Study results that report hospital charges have been transformed to costs using cost-to-charge ratios, a well-established method in the literature.²⁸ All costs are reported in 2015 dollar amounts and on a per-HAC-case basis. For example, the estimate of $5,746 for ADE means that for each ADE, on average, the hospital incurs an additional $5,746 in costs caring for that patient above and beyond the costs associated with an inpatient stay for the same patient without an ADE.

Exhibit 7.  Summary of meta-analysis additional cost estimates

More robust literature and higher overall additional costs were found for infectious HACs compared to non-infectious HACs. On average, 6.4 studies were included in estimates for infectious HACs, whereas only an average of three studies were available for non-infectious HACs. Of the infectious HACs (CAUTI, CLABSI, SSI, VAP, and CDI), we found the average cost attributable on a per-case basis to be approximately $31,000. The least expensive infectious HAC is CAUTI ($13,793), and the most expensive is CLABSI ($48,108), although both had wide ranges of estimates in the literature and wide confidence intervals in our results. The estimated attributable costs of non-infectious HACs are generally less than those of infectious HACs, with only VTE ($17,367) and pressure ulcers ($14,506) generating an estimated additional cost in the range of the infectious HACs.

To estimate excess mortality, we combined the results of meta-analysis with estimates of underlying mortality in the population as shown in Exhibit 8. For each HAC, we list the number of studies, the range of relative risk of death estimates from those studies, our pooled meta-analysis relative risk estimate, underlying mortality in the population, and finally, excess mortality and 95% CI for whether the estimated excess mortality is statistically different from zero. Underlying mortality values were taken from the literature and reflect our best estimate of the mortality rate for the population at risk for each of the HACs. More details on underlying mortality, including sources for each estimate, can be found in Appendix B.

Excess mortality is defined as the number of additional deaths due to a given HAC and shown as the number of deaths per HAC case. For example, we estimate for CLABSI that there are 0.15 excess deaths for each case. Stated another way, for every 1,000 CLABSI cases there are 150 excess deaths. Excess mortality is calculated as the difference between the probability of death among those who have the HAC and the probability of death among those who do not have the HAC but are at risk. The formula used to estimate excess risk and sources for underlying mortality estimates are provided in Appendix B.

Exhibit 8. Summary of meta-analysis excess mortality estimates

^*No studies could be used in our relative risk-based meta-analysis methods, so estimates were produced from an alternative method described in more detail in the OBAE section below.  

Our lowest estimates of excess mortality per case of HAC were for ADE, CAUTI, and SSI. It is possible that these HACs have been met with extensive in-hospital tracking and prevention efforts, compared to the other HACs in our study. The HACs with the highest excess mortality were CLABSI and VAP. These HACs tend to occur in sicker populations with an already increased risk for mortality. It is possible that the relatively higher risk of mortality from these HACs is due in part to the underlying morbidity in the types of populations vulnerable to such conditions. Estimates for VAP were extracted from a body of literature that was diverse in the types of specialty populations studied and thus may have limited generalizability. Of note, CLABSI has the highest estimate for both additional cost and excess mortality.

HAC Specific Considerations

The volume of literature, quality of studies, and relevance to our objectives varied for each of the HACs investigated. In this section, we briefly discuss these considerations for each HAC. Some of the factors we considered included:

• Recency of the data.
• Patient population studied related to the general population at risk for the HAC.
• Methods used to calculate cost and/or mortality.
• Analytic strategies used to assess additional cost and/or excess mortality.
• Generalizability of patient population studied to U.S. populations.
• Differences in HAC definitions used by each study.
• Variability in individual study estimates.

More details on each of the studies included in each estimate are provided in Appendix D. Forest plots for each additional cost and excess mortality meta-analysis for each HAC can be found in Appendix E.

Adverse Drug Events

Based on two studies reporting cost data, we estimated the additional cost for hospital-acquired ADE to be $5,746 (95% CI: -$3,950 to $15,441), whereas excess mortality, based on six studies, was estimated at 0.012 (95% CI: 0.003 to 0.025) per HAC case (meaning for every 1,000 in-hospital ADE cases, there are 12 excess deaths). The datasets used by our set of articles for these estimates are mixed. Two used hospital or administrative data; two used data from Can Rapid Risk Stratification of Unstable Angina Patients Suppress Adverse Outcomes with Early Implementation of the ACC/AHA Guidelines (CRUSADE),³⁴ a registry of patients with unstable angina receiving antithrombotic agents; one used the HCUP-NIS; and the last used MPSMS data.^35,36,37,38 This mix of data resulted in a mix of patient populations studied—from all Medicare beneficiaries to adult surgical or cardiology patients—and methods used to identify HAC cases, from ICD-9 codes to reported dosing and surveillance.

The confidence interval for our additional cost estimate overlaps zero, indicating that the plausible range for the true estimate (with 95% confidence) includes no additional costs related to ADE. We believe this is due to the low number of articles we were able to include. Another caution: we were only able to include studies involving the administration of opioids and, thus, this estimate may not be generalizable to anticoagulants, hypoglycemic agents, or adverse drug events involving other drug classes.

For our mortality analysis, we were able to capture results from six studies collectively dealing with the three drug classes in the Common Formats for Surveillance definition of ADEs, including opioids, anticoagulants, and hypoglycemic agents. There was significant variability in reported mortality between studies, due mainly to variation in patient population and data source. For example, one study that focused on Medicare beneficiaries and multiple drug classes reported a relative risk more than double that of studies involving all adults and one drug class.³⁵

Catheter-Associated Urinary Tract Infections

Based on six studies reporting cost data, we estimated the additional cost for hospital-acquired CAUTI to be $13,793 (95% CI: $5,019 to $22,568), whereas excess mortality, based on four studies, was estimated at 0.036 (95% CI: 0.004 to 0.079) per HAC case (meaning for every 1,000 in-hospital CAUTI cases, there are 36 excess deaths). All patients in the considered studies were adult (ages >18) except for one, which studied pediatric patients exclusively (age 1–17 years)³⁹ Although all studies reported sample size either in terms of number of cases or number of patients with CAUTI, we found wide ranges of sample sizes from 18 in 6-year pooled data on colorectal resection patients⁴⁰ to 105,113 in 10-year pooled data on surgical oncology patients.⁴¹ These discrepancies were largely due to the source of data and definition of populations used in each study. For instance, some studies involved all inpatient populations,^37,42 yet the majority of studies focused on specific medical and/or surgical conditions (e.g., surgical oncology in Sammon 2013; colorectal resection in Byrn 2015).^40,41 Additionally, the scope of the studies varied, from hospitals in a single network using data from EMRs^40,42 to nationally representative samples, with four studies using HCUP-NIS,^39,41,43,44 one using MedPAR claims,⁴⁵ and one using Cardinal Health MedMined data.⁴⁶ These factors potentially influenced our cost and mortality estimates, as exhibited in the large variations in individual estimates.

Central Line-Associated Bloodstream Infections

Based on seven studies reporting cost data, we estimated the additional cost for hospital-acquired CLABSI to be $48,108 (95% CI: $27,232 to $68,983), whereas excess mortality, based on five studies, was estimated at 0.15 (95% CI: 0.070 to 0.027) per HAC case (meaning for every 1,000 in-hospital CLABSI cases, there are 150 excess deaths). Individual studies produced a broad range of cost estimates for CLABSI, ranging from $18,000 to more than $90,000. The study with the lowest cost estimate was also the most recent study, using data from 2006 through 2012.⁴² Overall, most of the studies included in meta-analysis focused on specific patient subpopulations including pediatric patients, intensive care unit patients, and those with specific conditions (e.g., epilepsy, cancer). Two of the cost studies and two mortality studies used national databases (HCUP-NIS); however, all four of these focused on specific patient subpopulations for analysis. Only studies of single hospitals or local hospital networks reported consequences of CLABSI for a general inpatient population.

Mortality studies used slightly different definitions of CLABSI. Two studies used lab results, and one used the CDC-NHSN to define cases.^42,47,48 CLABSI definitions used in cost studies also varied from clinical surveillance criteria to ICD-9-based definitions; however, these differences did not seem to influence the resulting attributable cost estimates.

Falls

Based on three studies reporting cost data, we estimated the additional cost for hospital-acquired falls to be $6,694 (95% CI: -$1,277 to $14,665), while excess mortality, based on one study, was estimated at 0.050 (95% CI: 0.035 to 0.070) per HAC case (meaning for every 1,000 falls cases, there are 50 excess deaths). Our search for recent literature on in-hospital falls in the United States returned very few results dealing with cost and/or mortality specifically for in-hospital events.  Even fewer studies provided outcomes that could be incorporated into our meta-analysis. Much of the literature, which was screened out of our meta-analysis, dealt with one of two topics: 1) studies reported on cost and/or mortality resulting from admissions due to a fall in the community and 2) articles studied the impact of fall-prevention programs and protocols on the incidence of in-hospital falls.

The three studies involved in our estimate of additional costs use either a national sample of orthopedic surgery patients or a small sample of adult inpatients specific to a hospital. This variation in sample size and type led to a wide range in initial cost estimates from the literature ($2,680 through $15,491). One study (by Bates, et.al) used for our cost estimate employed data from the late 1980s to the early 1990s.⁴⁹ In addition, many more recent studies that address additional cost due to in-hospital falls base their costs calculations on the Bates article.⁴⁹ Costs associated with prevention efforts, as well as direct and long-term costs of care after a fall that requires hospitalization, have been measured but are outside the scope of this analysis.^50,51,52

The excess mortality estimate only represents the results of one study that used raw numbers found in the HCUP-NIS and should be treated with caution.⁵³ The dearth of literature on excess mortality may be due to the difficulty in finding reliable sources of data on injuries linked exclusively to in-hospital falls. Data on falls related to other health care settings, such as nursing homes, were not included in our analysis. While this may be a more relevant source of outcomes for falls specifically, based on articles found during screening, literature containing usable cost and mortality data is still limited for these other settings as well and may still face the same limitation of differentiating between falls leading to an admission and falls occurring in the institution.

Obstetric Adverse Events

Based on two studies reporting cost data, we estimated the additional cost for hospital-acquired OBAE to be $602 (95% CI: -$578 to $1,782). Our estimate of additional costs associated with OBAE is based on two studies reporting on a subset of conditions included in the maternal adverse event definition. One study used national representative data (HCUP-NIS) for obstetrical trauma (defined based on ´óÏóÊÓƵPSI).⁵⁴ The other study used 2010 all deliveries in a single State, and adverse events included postpartum hemorrhage, preeclampsia/eclampsia, and anesthesia-related adverse events.⁵⁵ With only two costs estimates that spanned a wide range ($13 to $1,190), the confidence interval for overall meta-analysis overlapped with zero increased costs. This additional cost estimate should be used with caution because it does not include a comprehensive set of maternal adverse events. For example, no data on costs associated with infections were found, and such infections could be costly.

Our systematic literature review found that there is a gap in current literature to examine the impact of maternal adverse events on hospital mortality in the United States. Instead of reporting on mortality associated with maternal adverse events, most studies analyzed maternal adverse events as the end point. This left us with only one study examining the risk of maternal mortality for adverse events acquired in hospitals, and the adverse event was obstetrical trauma only.⁵⁴ Further, this study found no increase in mortality for obstetrical trauma.

In addition, given the low incidence rate of maternal adverse events (1 percent) and low maternal mortality rate (0.02 percent), data analysis would require a national or a combination of multi-state databases across multiple years in order to achieve a large enough sample size to detect any increased risk. Furthermore, databases such as HCUP-NIS may have limitations to clearly identify cases (e.g., some researchers stated that they cannot distinguish a condition acquired prior to or during hospitalization). A surveillance system for maternal adverse events (not only for maternal mortality) would be helpful to understand the relationship between the adverse events and the associated outcome, including mortality and resource utilization.

Since we were not able to identify studies providing estimates of mortality due to OBAE, we used an alternative method to directly estimate the excess mortality using data on incidence of maternal adverse events and the risk of death among women experiencing maternal adverse events.

First, we estimated the total number of deaths due to maternal adverse events based on National Vital Statistics Data and CDC Pregnancy Mortality Surveillance Systems. From these sources and published literature, we estimated:

• Total number of live-births: 3,978,497 (2015 data).⁵⁶
• Overall maternal mortality rate: 23.8 per 100,000 live births (2014 data).⁵⁷
• Proportion of overall maternal deaths related to pregnancy: 38.2 percent (2011-2013 data).⁵⁸
• Percentage of pregnancy-related deaths due to adverse events: 37.1 percent (2011-2013 data).⁵⁸

We included the following conditions as the adverse events: infection (12.7 percent), hemorrhage (11.4 percent), hypertensive disorders of pregnancy (7.4 percent), amniotic fluid embolisms (5.5 percent), and anesthesia complications (0.1 percent). From the product of combining the number of live births, the maternal mortality rate, the maternal mortality rate related to pregnancy, and the percent of pregnancy-related deaths due to adverse events, we estimated the annual number of inpatient deaths due to maternal adverse events in the United States was 134, based on the assumption that all of the OBAE-related deaths happen in the inpatient setting.

Second, we estimated the incidence of maternal adverse events during delivery hospitalizations from four nationwide studies.^59,60,61,62 The incidence rate ranged from 220 to 1,148 per 100,000 deliveries. The studies varied by sample size (116,000 to 49 million), study period (1991 through 2011), study duration (3 years to 11 years), and data sources (one used National Institute of Child Health and Human Development Maternal-Fetal Medicine Units Network, one used National Hospital Discharge Survey, and two used HCUP data). They also differed in the conditions that were counted as maternal adverse events (all four studies included infection/sepsis, hemorrhage/blood transfusion, and eclampsia/hypertensive complications; three studies included amniotic fluid embolisms and anesthesia complications; two studies included intracranial injuries and internal injuries of thorax, abdomen, and pelvis; one study included iatrogenic events. Among those conditions, hemorrhage/blood transfusion is the most commonly acquired condition (137 to 1,044 per 100,000 deliveries), followed by eclampsia/hypertensive complications (48 to 63 per 100,000 deliveries), and infection/sepsis (17 to 33 per 100,000 deliveries). We used meta-analysis to estimate the overall incidence rate for maternal adverse event as 688 (95% CI: 257 to1,118) per 100,000 deliveries. Using a total of 3,978,497 live-births in the U.S. in 2015, the total number of cases with maternal adverse events were estimated as: 27,372 (95% CI: 10,225 to 44,480).

Dividing the number of pregnancy-related maternal deaths due to adverse events calculated in step one, by the total number of maternal adverse events calculated in step two, we arrive at an estimate of 0.005 (95% CI: 0.003 to 0.013) for excess mortality due to OBAE (meaning for every 1,000 OBAE cases, there are 5 excess deaths).

From these numbers it is also possible to calculate the percentage of maternal inpatient deaths due to OBAE. For this calculation, we took the number of inpatient deaths due to maternal adverse events (i.e., 134 as calculated earlier) and divided by the total number of inpatient maternal deaths in the United States. Total inpatient maternal deaths is calculated from the total number of live births, overall mortality rate, and percentage of maternal mortality occurring in the inpatient setting (based on the literature, estimated to be 62 percent).⁶³ Thus, we estimated 23 percent of maternal inpatient deaths are due to OBAE. Of note, this calculation assumes all OBAE-related deaths happen in the hospital setting, which, if not true, would mean excess mortality and percentage of inpatient maternal deaths due to OBAE are overestimates.

Pressure Ulcers

Based on three studies reporting cost data, we estimated the additional cost for hospital-acquired pressure ulcers to be $14,506 (95% CI: -$12,313 to $41,326), whereas excess mortality, based on three studies, was estimated at 0.041 (95% CI: 0.013 to 0.093) per HAC case (meaning for every 1,000 pressure ulcer cases, there are 41 excess deaths).

We identified six studies providing estimates of costs and/or mortality for hospital-acquired pressure ulcers. Patients in the included studies were mostly adult (ages >18), except for one that studied pediatric patients exclusively (ages 1–17 years),³⁹ and another that studied patients of all ages, including patients younger than 18 years of age.⁵⁴ All studies used nationwide data with five relying on HCUP-NIS.

The sample size and standard error of individual study estimates ranged widely from 148 patients across 4 years³⁹ to 670,767 patients across a 5-year study period.⁶⁴ These differences were largely due to the source of data and the definition of population applied in each study and likely are the cause of the large confidence interval for our additional cost estimate (-$12,313 to $41,326). For instance, some studies involved all inpatient populations, whereas others focused on specific medical and/or surgical conditions (e.g., epilepsy in Mendizabal 2016, and surgical patients in Spector 2016).^65,66 Given the evidence that the incidence of pressure ulcers increases with age,⁶⁴ we performed a sensitivity analysis that excluded the pediatric study (Goudie 2015) and estimated the additional cost for hospital-acquired pressure ulcers among adult inpatients to be $12,712 ($278 to $25,145).

Surgical Site Infections

Based on five studies reporting cost data, we estimated the additional cost for hospital-acquired SSI to be $28,219 (95% CI: $18,237 to $38,202), whereas excess mortality, based on three studies, was estimated at 0.026 (95% CI: 0.009 to 0.059) per HAC case (meaning for every 1,000 SSI cases, there are 26 excess deaths).

Only two studies were explicit about the types of infection included in their definition of SSI. One study counted all superficial, deep, and organ-space SSIs, whereas the other included only deep and organ-space infections.^67,68 The data sources used and cost estimates found in both of these studies did not vary from those in the other included studies. Three studies were regionally specific and involved single hospitals, one of which included all surgical patients that met the HAC definition.^42,67,68 The three remaining studies used national claims databases but focused on specialized surgical populations.^41,69,70

Studies for both cost and mortality estimation had a wide range of individual estimates. Studies varied in their source of cost data, from national claims databases (i.e., HCUP-NIS) to hospital administrative data systems. For example, the study with the lowest cost estimate ($11,778) involved chart review at a single medical center and had one of the lowest number of cases (N=186) among all included studies.⁶⁷ For mortality, studies with the largest and smallest estimates (relative risk of 6.18 and 1.79, respectively) both used close to 10 years of data from the HCUP-NIS database, but the study with the smallest estimate included a much larger population of surgical patients.^41,69

Ventilator Associated Pneumonia

Based on five studies reporting cost data, we estimated the additional cost for hospital-acquired VAP to be $47,238 (95% CI: $21,890 to $72,587), whereas excess mortality, based on 10 studies, was estimated at 0.14 (95% CI: -0.11 to 0.73) per HAC case (meaning for every 1,000 VAP cases, there are 140 excess deaths).

Compared to the other HACs studied, the cost literature for VAP is older. The most recent estimate comes from 2009, and two others date from prior to 2000.^71,72,73 The two studies prior to 2000 reported the lowest attributable costs at $19,000 and $33,000. All of the later studies’ estimates fall between $40,000 and $80,000.

The majority of VAP studies included in both cost and mortality estimates were conducted among ICU patients with only three studying patients outside of these units, one looking at cancer patients, and another examining all hospitalized patients.^41,74,75 Two of the studies included in the mortality estimate reported on pediatric populations, one from a PICU and the other a NICU.^76,77 Most VAP studies drew data from hospital medical records or databases that combined records from several hospitals. This enabled them to use VAP definitions that incorporated clinical information such as laboratory testing that closely mirror the QSRS definitions. Because many of these studies were conducted in single institutions or small groups of hospitals, the number of cases was small in each study. The preponderance of studies reporting on local data may limit the generalizability of estimates to the entire United States. 

More than half of the studies included in the mortality estimate used regression modeling techniques to estimate mortality due to VAP; however, there were a large minority (four studies) reporting only deaths for VAP patients and a matched comparison group. Estimates of relative risk for mortality varied and included two studies showing protective effects for ventilated patients with pneumonia compared to those without pneumonia.^71,74 When these studies are excluded, the estimate for excess mortality increases to 26 percent of cases (meaning that for every 1,000 cases there are 260 excess deaths).

In 2013, the CDC introduced the concept of ventilator–associated events (VAEs).⁷⁸ This represents a fundamental shift in the focus of HACs related to mechanical ventilation from a single adverse event (i.e., pneumonia) to a broader concept that includes additional potential pathophysiologic etiologies (e.g., VTE, volume overload, non-pulmonary infections).⁷⁹ While VAEs have a significant impact on patient outcomes (e.g., mortality and hospital length of stay), much work remains to further characterize the effect of VAEs on cost and mortality in isolation and in relation to other HACs.⁸⁰

Venous Thromboembolism

Based on four studies reporting cost data, we estimated the additional cost for hospital-acquired VTE to be $17,367 (95% CI: $11,837 to $22,898), whereas excess mortality, based on nine studies, was estimated at 0.043 (95% CI: 0.040 to 0.078) per HAC case (meaning for every 1,000 VTE cases, there are 43 excess deaths).

Eleven studies were included in our review for VTE. Four addressed costs related to VTE leading to a pooled estimate of $17,367 additional costs for each VTE event. Most of the studies included in our estimates reported on adult inpatient populations,^{65,69,81,82,83,84,85} with two studies considering patients of all ages^86,87 and one study focused on pediatric patients exclusively.³⁹ Sample size varied greatly in terms of total population, number of cases identified, and incidence rate of events. These variations were largely due to the definition of population applied in individual studies. For instance, some studies evaluated all inpatient populations, yet others focused on specific medical and/or surgical conditions (e.g., epilepsy in Mendizabal 2016; cirrhosis as denominator condition for Ali 2011; patients who underwent an ablative procedure for a malignant oral cavity, laryngeal, hypopharyngeal, or oropharyngeal neoplasm for Hennesey 2012).^65,82,83

Clostridium difficile Infections

Based on nine studies reporting cost data, we estimated the additional cost for hospital acquired CDI to be $17,260 (95% CI: $9,341 to $25,180), whereas excess mortality, based on 13 studies, was estimated at 0.044 (95% CI: 0.028 to 0.064) per HAC case (meaning for every 1,000 in-hospital CDI cases, there are 44 excess deaths). Methods of the included studies ranged from analysis of national hospital discharge data to reviews of a single hospital’s CDI rates. The majority of studies in our analysis focused on specific patient subpopulations (e.g., trauma patients, cancer patients, those admitted for organ transplant). Analytic methods also varied considerably, as studies using a matched control group tended to be more comparable to cases on observed covariates and may be a better approximation of attributable cost and/or mortality than those studies using a pooled control sample. Finally, few studies used clinical definitions of C. difficile infection and instead relied on ICD-9-based definitions, which may miss cases and may misclassify community-acquired cases as hospital-acquired.