In recent years, increased attention has developed concerning improvement of hospital and health care outcomes in the United States. This development has resulted from a combination of factors related to inpatient complications, hospital readmissions, and other indicators.

A major cause of this development has been research that demonstrates the relationship between adverse health care outcomes and higher costs to providers and payers. Studies have demonstrated that patients with inpatient complications such as pneumonia, urinary tract infection, and clostridium difficile colitis have much longer hospital stays and related labor and pharmaceutical costs than those who do not [1-3].

Related to these costs is a new urgency to reduce health care expenses in society. In the United States, all major health care payers, including Medicare, Medicaid, and private insurance, are under a large amount of pressure to reduce spending. This situation strongly suggests that current increases in these expenditure can no longer be sustained [4,5]. The potential to control these costs through financial penalties for adverse outcomes holds the potential to reduce health care spending while improving patient care [6].

In addition to these factors, efforts to improve health care outcomes are benefiting from the development of electronic software for analysis of patient specific data. Tools such as the Potentially Preventable Complications system developed by 3M Health Information Systems can analyze large amounts of administrative data at the aggregate and patient specific levels. The information produced by these tools can guide clinical management initiatives [7,8].

This study described the use of administrative data and one of these algorithms to support reduction of inpatient complications in a large urban hospital in Syracuse, New York during a four year period. It demonstrated the use of this software to identify and address adverse outcomes at the aggregate and specific levels.

This study involved the use of administrative data to identify and manage inpatient complications in St. Joseph’s Hospital Health Center, in Syracuse, New York during a four year period. This hospital has the largest inpatient volume (23,832 inpatient discharges excluding well newborns in 2011) of the acute care facilities in Syracuse. Other hospitals in Syracuse that participated in the program included Crouse Hospital (20,540 discharges) and Upstate University Hospital at Community General (6,959 discharges) (Hospital Executive Council, Unpublished data, 2012).

The program involved the staffs of St. Joseph’s Hospital Health Center and the Hospital Executive Council, the cooperative planning organization for the Syracuse Hospitals. Historically, the Syracuse Hospitals have worked through the Council to improve health care efficiency and outcomes in Central New York [9].

The study was developed as part of a demonstration program including the Hospital Executive Council, St. Joseph’s Hospital Health Center, and 3M Health Information Systems. This program involved the use of the 3M Potentially Preventable Complications software and hospital administrative data to identify and manage inpatient complications.

The Potentially Preventable Complications software includes extensive logic for identifying inpatient hospital complications in administrative data. At the summary level, it identifies complications at the patient specific level based on specific secondary diagnoses. These diagnoses, which are identified by the reporting indicator as Not Present on Admission, are assumed to be candidates for inpatient complications. The software then screens these diagnoses with specific exclusion criteria such as logical sequelae of principal diagnoses and services such as trauma and neonatal care for which inpatient complications have not been clearly identified.

In the demonstration program with 3M Health Information Systems, the Hospital Executive Council generated Potentially Preventable Complications data for St. Joseph’s Hospital Health Center each month beginning in 2009. These data included patient specific spreadsheets and summary tables for PPCs that were the subject of clinical interventions at the Hospital.

In response to this information, the Hospital staff identified whether patient records confirmed that each patient experienced the complication(s) identified by the software and whether they received the clinical intervenetions designed to avoid these adverse outcomes. This information was used to monitor development of the program by the Hospital Executive Council and for internal follow up by the hospital.

The charts of patients that were identified as having a particular potentially preventable complication were reviewed in detail to identify commonalities. The first intervention was to determine whether the documenttation supported the diagnosis that resulted in the PPC. Coding errors had to be eliminated before clinical interventions could be put in place. This gave the hospital a more accurate picture of the scope of the problem. This was accomplished by reviewing each clinical record and by comparing them to coding summaries and coding definitions [10].

The charts fell into two distinct categories: The first were records where the documentation did not support the coding. The other category included charts where the clinical information did not support the diagnosis. By educating coders about errors and common mistakes, coding errors were systematically eliminated. By educating providers about how diagnosis of pneumonia and urinary tract infections were coded, their documentation regarding these complications became clearer.

With administrative “slack” taken out of the system the true clinical situations could be identified and mitigated. Over 100 chart reviews were conducted to define what specific outcome issues there was at the hospital. The staff wanted to be as specific as possible for the patient populations so that they could effectively deploy resources to reduce the complications of interest. The hospital focused on four complications: Pneumonia, Urinary Tract Infection, Clostridium Difficile Colitis, and Decubitus Ulcers.

Charts were examined to identify if patients were receiving the basic care documented in the literature to prevent hospital acquired pneumonia [11]. Specifically, nursing charting was reviewed to determine if the patients with the complication received mouth care, had the use of incentive spirometry documented, were out of bed ambulating, and if they had the head of their bed elevated [12,13]. It was clear from chart reviews that incentive spirometry and mouth care were opportunities to improve care. Units and services with the highest numbers of cases were selected for education regarding incentive spirometry use and documentation. Daily rounding was instituted to ensure that nurses were reminding patients to use their incentive spirometers and that they documented its use.

The data from each source were imported into a Microsoft Access database and compared using medical record number, account number and admission date. Records for review were then sorted by comparing them with valid culture results to arrive at the study group. Chart reviews were conducted to determine what care was documented. These reviews included the collection of the infection indicators of white blood cell count and temperature associated with culture collection. The reviews determined for each case if there was a urinary catheter placed during the patients admission, how many days it was in place, and the number of times care was charted in association with it. For those infections where a urinary catheter was in place, the number of catheter days was compared with the number of times catheter care was charted to approximate a rate of care episodes per catheter day.

The key intervention in reducing hospital acquired urinary tract infection was to reduce the overall usage of urinary catheters. To do this, the problem was addressed by a number of approaches. First, using the CDC’s guidelines for reducing catheter associated urinary tract infections, standard indications for catheter insertion were implemented [14]. Three interventions were effecttive in reducing the number of catheter days per patient (the in process measure). Firstly, an RN driven foley removal protocol for inpatients was instituted. Essentially, if the indications no longer existed, the catheter was removed automatically by the nurse. Secondly, for surgical patients, an education program for physician assistants, residents and nurse practitioners about the correct indications for indwelling catheters was instituted. Finally, it was determined that the majority of urinary catheters were inserted in the emergency department. As a result, an education program for the both the nurse and providers in the emergency department about the indications, and effects of indwelling catheters for patients during the entire hospitalization was carried out. All of these interventions resulted in approximately a 20 percent reduction of catheter days per patients and a precipitous decline in hospital acquired urinary tract infections.

Cases identified by the 3M Potentially Preventable Complication (PPC) software were compared with those reported to the National Health Safety Network (NHSN) at the Centers for Disease Control and Prevention (CDC) [15-17]. They were also cross referenced with clostridium difficile colitis cases identified by Care fusion/Med mined Virtual Surveillance Indicators. After cases were identified and cross referenced, a total of 124 unique records identified as hospital onset were reviewed in detail using data found in the clinical documentation system, the orders system and the pharmacy system. Finally, patient room assignments were reviewed.

The initial review was carried out using the pharmacy system, Horizon Meds Management (HMM). The reviewer reviewed notations made by pharmacist regarding antibiotic prescribing and indications. This documentation is not part of the patient’s clinical record. The reviewer attempted to identify the patient’s original infection and the antibiotics used to treat it. When this was not documented in HMM, the clinical record was reviewed.

The findings of infection requiring antibiotic treatment are listed in Table 1.

Next the antibiotics used to treat these infections were review and counted. Vancomycin and Flagyl were excluded from this count. The results are listed in Table 2.

It is worthwhile to know that 84 percent of patients received at least one of the antibiotics listed in Table 2 and 45 percent of patients received at least two. These findings are consistent with the findings noted by Pepin et al. in their study “Emergence of Fluotoquinolones as the predominant risk factor for Clostridium DifficileAssociated Diarrhea: A cohort study during an epidemic in Quebec” [18].

It was clear from an extensive review of patient records and careful comparisons of other data sets that a previously identified quality improvement opportunity in the care of the decubitus ulcer PPC was far more theoretical

than practical. Coding errors, careful nursing documentation, poor initial patient condition and co-morbidities and additional PPCs have conspired to negate any real financial gain to be had from further improvements to decubitus care alone.

Analysis of the data involved the impact of interventions for individual inpatient complications, as well as the development of summary tables including numbers of complications, at risk populations, and rates per 1000 discharges for St. Joseph’s Hospital Health Center. In order to define the full and most recent impacts of the program, one analysis included data for January-December 2008-2011, while the other included data for JanuaryMarch 2008-2012. In both analyses, data for individual and aggregate complications were identified.

The first component of the results involved the individual complications addressed by the project. Data concerning all individual complications are summarized in Tables 3 and 4 which follow.

With respect to pneumonia, increased incentive spirometry use and documentation resulted in the direct reduction of hospital acquired pneumonia. Changes in complication rates for pneumonia are identified in Tables 3 and 4 which follow.

With respect to urinary tract infection, the association between urinary catheters and urinary tract infections was examined. The review looked at all of these infections, as well as their relationship to indwelling urinary catheters. It found that 78% (n-76) of St. Joseph’s hospital acquired urinary tract infections were in patients who had been catheterized during their admission. On average, a patient who developed a urinary tract infection had an indwelling catheter for 12 days. St. Joseph’s standard for catheter care is that it be performed and documented once daily. The review found that only 25 percent (n-19) of patients with positive urine cultures received this care correctly. In 34 percent of the cases the patient received less than what was required (n-23) and curiously, 45 percent (n-34) of the patients received too much care, that is to say they had more episodes of care then recommended. Changes in complication rates for urinary tract infection are identified in Tables 3 and 4 which follow.

With respect to clostridium difficile colitis, another factor examined was the timeliness of isolation precautions associated with the onset of symptoms. To determine this all of the nursing orders for “enteric” and “contact precautions C. difficile” were compared with orders for “Stool for C. difficile”. The basic presumption of this comparison was that for a “Stool for C. difficile” to be ordered there was reason to believe that the patient is in some way symptomatic of the disease. As such, if there was a reasonable expectation that a patient is symptommatic, they were placed isolation in accordance with the hospitals isolation manual and the CDC’s current best practice guidelines. The examination included all inpatients who were placed on isolation and all of those who had an order for stool for C. diff. This produced 607 unique order combinations for inpatients that were admitted for at least 48 hours. On average, patients who were placed on isolation in greater than one hour waited five hours for the isolation order to be entered.

Finally, an examination of positive clostridium difficile colitis results was made, irrespective of the onset to determine if any other environmental factors could be identified. This review specifically looked at the rooms patients with positive results stayed in. It found that 15 patient rooms (7 percent of rooms) accounted for 54 (20 percent) of positive results. This was significant in that environmental contamination with C. Diff spores is a major contributing factor in developing the disease in patients with other risk factors. Changes in complication rates for clostridium difficile colitis are identified in Tables 3 and 4 which follow.

With respect to decubitus ulcers, prolonged length of stay seemed to contribute to the development of the complication. In other words, the ulcer resulted from the extended stay, rather than the long stay resulting from the ulcer. Changes in complication rates for decubitus ulcer are identified in Tables 3 and 4 which follow.

The second component of the results involved Potentially Preventable Complications at the aggregate and diagnosis specific levels for January-December 2008-2011. Relevant data are summarized in Table 3.

This information demonstrates that, at the aggregate level, the PPC rate per 1000 discharges at the hospital declined by 33.4 percent, from 56.11 to 37.37 during this period. This occurred as the number of PPCs declined by 13.4 percent from 1,035 to 896, while the at risk population increased by 30.0 percent, from 18,446 to 23,975. It was notable that the hospital was able to reduce complications at a time when its inpatient population was increasing substantially.

At the PPC specific level, the principal drivers of the decline were high volume diagnoses that were addressed by specific interventions. For pneumonia (PPC 04) the rate declined by 45.7 percent from 14.07 to 7.64 per 1,000 discharges between 2008 and 2011. Most of this reduction occurred between 2009 and 2011. For urinary tract infection (PPC 16), the rate declined by 23.7 percent from 8.37 to 6.39 between 2008 and 2011. As a result of an increase between 2008 and 2009, all of the decline occurred during the last three years.

Hospital Executive Council—3M Health Information Systems PPC demonstration program.

Hospital Executive Council—3M Health Information Systems PPC demonstration program.

The lower volume complications that were addressed by specific interventions contributed less to the overall PPC decline. The rate for clostridium difficile colitis increased from 1.60 to 3.98 between 2008 and 2011. The rate for decubitus ulcer declined from 3.02 to 2.04.

In addition to pneumonia and urinary tract infection, PPCs that were not addressed by specific interventions also contributed to declines in the overall rate. These included pulmonary embolism (PPC 03), ventricular fibrillation (PPC 10), and septicemia (PPC 20).

In order to provide more updated information concerning Potentially Preventable Complications at St. Joseph’s Hospital Health Center, the third component of the study included information for January-March 2008- 2012. These data are summarized in Table 4.

This information demonstrated that between JanuaryMarch 2008 and 2012, the aggregate PPC rate at the hospital declined by 48.1 percent, from 58.51 to 30.38 per 1,000 discharges. This decline progressed throughout the five year period. Between the two most recent periods, January-March 2011 and 2012, the reduction was 29.2 percent, from 42.88 to 30.38 per 1000 discharges.

As in the annual data, high volume PPCs that were addressed by interventions were major drivers of the aggregate reduction. These included pneumonia (PPC 04), where the rate declined by 51.7 percent, from 14.46 to 6.99 per 1000 discharges and urinary tract infection (PPC 16) where the rate declined by 14.4 percent, from 7.72 to 6.61 per 1000 discharges, during the five year period. It was notable that the rates for both of these complications increased between January-March 2011 and 2012.

As in the annual data, smaller volume PPCs did not contribute greatly to the overall decline for this time period. The rate for clostridium difficile colitis (PPC 15) increased, while the rate for decubitus ulcer declined.

As in the annual data, additional PPCs that were not addressed by specific interventions also contributed to declines in the overall rates. These included pulmonary edema and respiratory failure (PPC 03), an 80.0 percent reduction; septicemia (PPC 20), a 72.3 percent reduction, and postoperative hemorrhage (PPC 23), an 80.7 percent reduction.

This study described the use of administrative data and computer software to support the reduction of inpatient complications in a large urban hospital during a four year period. It demonstrated how these resources could be employed to identify and improve these outcomes for a wide range of diagnoses.

The interventions to reduce complications implemented by St. Joseph’s Hospital Health Center were derived from research literature and local experience. Based on recommendations from published research, they were adapted to the needs and resources of the hospital. The results of the study demonstrated that the identification and use of these interventions were largely successful.

The experience of St. Joseph’s Hospital Health Center demonstrated how aggregate complications data could be used to identify address specific diagnoses for intervenetions. Using the Potentially Preventable Complications software, the hospital staff was able to select complications with relatively high volumes, such as pneumonia and urinary tract infections, that would have the largest impact on aggregate outcomes and related costs. The staff was also able to identify complications with lower volumes, such as clostridium difficile colitis and decubitus ulcer that were of interest.

The administrative data and computer software were useful in identifying patient specific issues with respect to documentation. The spreadsheets that were developed from these resources contributed to improvements in the coding of administrative data that clarified the actual numbers of complications that occurred. This process improved evaluation of hospital quality assurance efforts to address these outcomes, as well as the accuracy of administrative data being used by reporting agencies in the public area.

The administrative data and computer software also made it possible to identify patients who received the program inventions, but also experienced the complications. From this perspective, it provided information for evaluation of the impact of interventions over time on a patient specific basis.

Through these applications, the staff of St. Joseph’s Hospital Health Center was able to use these resources at both the patient specific and aggregate levels to improve care and reduce related costs. The aggregate data provided perspectives concerning this information across a wide range of individual complications and through total frequencies and rates.

In summary, this approach to improving patient outcomes was simple and direct. It provided a means of communicating and managing outcomes data that could be understood and used by a wide variety of health care providers.