| Tufts-New England Medical Center–Neurosurgery, Boston, Mass.
Correspondence to: Received: Feb. 24, 2007 Key Words: diagnosis-related group, healthcare, outcomes
assessment, pay for performance, quality Abbreviations |
The quality of healthcare delivered to patients has been a prominent issue for decades. One of the earliest databases was established in 1971 to monitor cardiac surgery patient outcomes with the goal of identifying measures that could improve patient care (14). However, substantial public and political attention was not focused on evaluating the quality of healthcare until after publication in 1999 of the Institute of Medicine report on medical errors. To Err is Human: Building a Safer Health System reported that 44,000 to 98,000 deaths of hospitalized patients each year were attributable to medical errors and that these deaths were preventable (20). Since this report, increased public awareness and political attention has resulted in the propagation of programs aimed at reducing medical errors, improving the quality of healthcare, and identifying potential indicators to measure quality. There also has been a marked increase in the last decade in the number of research awards and published articles related to patient safety (28).
The quality indicators currently identified by the AHRQ include mortality rates and types of procedures. This data is extrapolated from insurance claims. To address concerns about accountability and transparency, in many states hospital-based data and individual surgeon reports have been posted online and are accessible by the public. These reports also are used by some healthcare insurance companies for interinstitutional comparisons of quality delivered. These quality assessments are the foundation for pay-for-performance programs that already are implemented in many areas of medicine. Ultimately, these programs will expand to incorporate all fields of medicine, including neurosurgery. The AHRQ has identified several quality indicators that impact neurosurgical practice: area-level rate for laminectomy or spinal fusion, carotid endarterectomy, craniotomy mortality, and acute stroke mortality.
Recently, a Tufts-NEMC healthcare insurer reviewed the mortality rate of hospitalized patients admitted under the DRG code of acute stroke. The DRG code is generated in a process that begins at patient discharge or death. Medical records staff reviews the patient chart and compiles the list of ICD-9 codes that describe treatment during the admission. The codes are subsequently forwarded to the insurer and assigned a single DRG code. The insurer compared the Tufts-NEMC mortality rate for the DRG code of acute stroke with that of other academic institutions in Massachusetts and identified hospital mortality rates that were deemed outliers or showed a significant change in mortality rate from year to year. For acute stroke patients admitted to academic institutions in Massachusetts, Tufts-NEMC had the lowest mortality rate in 2003 but in 2004 had the second worst mortality rate in the state. Even though in 2004 the Tufts-NEMC mortality rate for this group of patients did not differ statistically from that of the cohort, the data prompted the institution to evaluate the mortalities of 2004 by internal audit.
| Abstract Pressure to reduce healthcare costs and improve quality of care has led some insurers to adopt a pay-for-performance system that incorporates quality indicators. Recently, an insurer evaluated the quality of Massachusetts hospitals using mortality rate for patients admitted under the DRG code of acute stroke. Based on the assumption that mortality is an outcome that reflects quality of care, the insurer identified a higher mortality rate for this patient group in 2004 at Tufts-NEMC than at most other academic centers in Massachusetts. The insurer also identified significant increase in mortality rate in this group from 2003 to 2004 at Tufts-NEMC, prompting an internal committee to review all charts for this patient group. The committee evaluated details of patient demographics, severity of illness, code status at the time of death, length of hospitalization, and hospital transfers and concluded that no significant contribution from medical error led to patient mortality. The committee found that simply reviewing DRG code data and in-hospital mortality rates was insufficient to accurately and reliably determine quality of healthcare for the admission diagnosis of acute stroke. |
Tufts-NEMC formed an internal committee to review the outcomes of acute stroke patients identified by the insurer for 2003 and 2004. The committee was composed of representatives from the departments of neurosurgery, neurology and internal medicine. All hospital charts from patients identified by the insurer with a principal DRG code of acute stroke who died in-hospital in 2003 and 2004 were reviewed and the data was compiled for comparison. Data collected included age, gender, primary diagnosis, co-morbidities, length of stay, code status at the time of death, and transfer status. Code status at the time of death was defined as DNR/DNI or CMO for palliative care. Charts were carefully reviewed for any iatrogenic factors, medical errors, or significant variations from the standard of care that may have contributed to mortality. An unpaired Student-t test was used to calculate statistical significance for some values.
Results
Mortality for patients admitted under the DRG code of acute stroke totaled
21 of 181 (11.6 percent) patients in 2003 and 31 of 204 (15.2 percent) patients
in 2004. As compared to the cohort of other academic hospitals in Massachusetts,
the lower mortality rate in 2003 was considered statistically significant (p
= 0.10), but the higher mortality rate in 2004 was not statistically different
from that of the cohort. However, the change in mortality rate from 11.6 percent
in 2003 to 15.2 percent in 2004 was identified by the insurer. During the 2003–2004
period, no significant changes were made in policy or protocol management that
might have affected the sample population.
Data from all 21 patient charts for 2003 were reviewed and compared with data from the 31 patient charts for 2004. Results are illustrated in Table 1. No significant difference existed between the mean age or gender of the patients. In each of the two years studied a similar number of patients presented either with neurological exams consistent with brain death or with such poor neurological condition that treatment was considered medically futile. These patients progressed rapidly to death within 24 hours of admission and were classified with length of stay of less than one day. For patients whose length of stay was greater than one day, hospitalization was significantly shorter in 2004 than in 2003 (4.7 versus 14.6, p = 0.013). Given that a few patients in 2003 had lengths of stay well beyond the mean (more than 60 days), which may have skewed the mean calculations, the median (10 in 2003 and three in 2004) and mode (three in each year) were calculated as well. A significant portion of the patients in each of the years was transferred from another acute care institution (66.7 percent in 2003 and 54.8 percent in 2004).
In 2003, at time of death all of the patients were designated CMO, DNR/DNI, or brain dead. In 2004, one patient died from a hemothorax as a complication from a central venous line placement; in two cases family discussions had been initiated toward CMO status; and the remaining 28 patients were CMO, DNR/DNI, or brain dead at the time of death.
The deaths of two patients in 2003 could be categorized as unexpected. One of these patients was incorrectly categorized by DRG code. The patient underwent a carotid endarterectomy and had a history of a stroke but did not have an acute cerebrovascular accident. The patient developed a myocardial infarction and bacteremia postoperatively and ultimately died after being designated CMO by the family. In the second case, a patient with a large intraparenchymal hemorrhage had an episode of pulseless electrical activity, and there may have been a delay in addressing it and calling a code. However this patient’s prognosis was very poor because of the size and location of the presenting intracerebral hematoma.
In 2004, there was one preventable death. This patient died from a hemothorax after a central venous line placement. The patient presented initially with a left frontal lobe hemorrhage from a cystic mass. An autopsy was not performed, but this patient likely had a tumor and thus should not have received the primary diagnosis of acute stroke at admission. Another patient in 2004 died of a complication, but the treatment given was appropriate. This patient was an elderly woman who developed a thalamic hemorrhage after receiving tissue plasminogen activator when she presented to the emergency department with hemiplegia and a negative head CT scan. Ultimately, the family converted the patient’s code status to DNR/DNI given her poor functional status, and the patient died from respiratory arrest secondary to aspiration pneumonia.
Overall, no significant differences other than length of stay were identified between the groups.
 |
With current national healthcare expenditures at a historic peak and future costs estimated to rise precipitously, cost-saving and quality-improving measures are becoming increasingly prevalent. Healthcare providers currently are leaning toward pay-for-performance systems in which physicians and institutions are rewarded for excellence of care. The goal of pay for performance is to reward quality of care by increasing reimbursement to institutions or physicians based on their performance, with the expectation that there will be decreases in long-term costs because care will be improved and complications will be prevented.
The pay-for-performance model is based on the ideas that the selected indicators accurately depict quality of care and that improved performance translates into decreased healthcare expenditures, but some of the quality indicators have not been thoroughly validated. Safavi summarizes many of the potential pitfalls with pay-for-performance systems when overemphasis is placed on measuring indicators at the detriment to quality of care (27).
In our series, the insurer grouped many differing pathological entities under the DRG code of acute stroke. These entities included diagnoses that have considerable variation in their prognoses, such as ischemic infarcts, intraparenchymal hemorrhages, acute subdural hemorrhages and aneurysmal subarachnoid hemorrhages.
Although some risk adjustment is incorporated, the current DRG codes do not convey a sense of illness severity. The grade of subarachnoid hemorrhage is not considered, which clearly has a significant impact on the prognosis (1, 11, 17–19). For example, DRG codes do not differentiate between patients who present with a subarachnoid hemorrhage of Hunt and Hess grade V or grade I. Although coding for additional comorbidities to some extent does convey the overall medical health of the patient, the neurological severity of each diagnosis is not represented in the current AHRQ system. This is problematic when comparing stroke mortality rates because a high percentage of the patients are transferred in moribund condition from other institutions. A change in referral patterns from year to year could significantly affect the number of patients in poor neurological condition being transferred to an academic medical center. Some other entities are attempting to incorporate measures that adjust for illness severity using a more specific coding system, but specific modifiers currently are not being applied to the DRG code of acute stroke.
The neurosurgical literature indicates a poor prognosis for patients with intracranial hemorrhages, bilateral fixed and dilated pupils, large left middle cerebral artery strokes, large dominant hemisphere intra-parenchymal hemorrhages, and Hunt and Hess grade V aneurysmal subarachnoid hemorrhages (1, 11, 13, 15, 17, 21–23, 25). Although overall survival rates can be increased by aggressive surgical interventions such as tracheostomy, gastric feeding tubes, and craniectomies for large strokes and hemorrhages, long-term functional status remains poor especially for elderly patients (1, 11, 13, 15, 22, 23, 25). Therefore aggressive measures such as tracheostomy and gastrostomy placement along with surgical decompression and craniectomies will improve outcomes as evaluated strictly by mortality but may not improve the quality of care or accurately reflect patient desires and family wishes. Also, aggressive treatments to decrease mortality rates while not improving functional outcomes may actually increase healthcare expenditures and the socioeconomic burden.
In attempting to reduce complication rates and avoid errors, regionalization of medical care has already been considered. In other surgical specialties, reports have documented a lower risk of complications with higher procedural volume (3–6, 7, 10, 12), but regionalization may have an associated increased cost per patient (3). Published quality evaluations may steer patients toward specialty centers, which preferentially treat high volumes of specific types of cases. In primary care, some centers have decided to focus patient management on complicated diabetic care whereas in neurosurgery, centralization of resources to treat diseases such as aneurysmal subarachnoid hemorrhages has occurred (3).
Although interinstitutional evaluations typically distinguish academic institutions from community hospitals, some discrepancies may exist within these designations. Institutions may be considered academic for specific specialties, but may not have residency training programs for other specialties. Furthermore, community institutions that do not have neurosurgical coverage are likely to transfer cases such as massive strokes and intracerebral hemorrhages to alternate institutions and therefore may have a biased lower mortality rate because higher risk patients are transferred to other facilities. In our series a significant portion of the mortality sample included transfers from other institutions (66.7 percent in 2003 and 54.8 percent in 2004).
Pressure to rank highly on public report cards may lead institutions to devise strategies that improve their outcome values while not necessarily improving quality of care. Notably, in the United Kingdom the exclusion of patients from the evaluated population was found to be the greatest predictor of quality achievement in primary care (9).
In fact, there are several ways for hospitals to lower their mortality rate for acute stroke DRG patients that have nothing to do with improved quality of care. One way is to have no emergency neurosurgery call coverage. Then all devastating cerebrovascular accidents must be transferred, resulting in a favorable stroke mortality rate at the referring institution. As mentioned previously, aggressive treatment of comatose patients with tracheostomies and gastrostomies would help decrease mortality but would be unlikely to improve their functional status or quality of care. Rapidly transferring CMO patients to hospice care, including to in-hospital hospice wards that are distinct from the institution itself, would avoid in-hospital mortality designation. Also, if patients with transient neurological symptoms such as transient ischemic attack are given an MRI scan, those with scans showing a possible tiny infarct can be coded with an admission diagnosis of stroke, resulting in an increased denominator for the complication rate and a decreased percentage of complications and mortality.
Other strategies may involve the under-reporting of errors or complications. Since publication of the IOM report on medical errors, significant research and effort has gone into reducing the number of adverse events in healthcare. Many studies have shown some improvement in reporting and reduction of errors after implementing systematic measures, but the reporting of errors carries with it a persistent stigma (16, 26). Many reports have stressed the importance of nonpunitive reporting for this system to function effectively (2, 16, 24), and others have advocated the need for tort reform to precede significant and accurate reporting of errors. Given the high number of malpractice lawsuits in the United States, a role in litigation for error reporting and quality rankings is sure to be found.
When Tufts-NEMC first received the report indicating that its DRG acute stroke mortality rate in 2004 was the second to worst in the state, we hoped that our internal review of medical records of patients who died would identify areas for improvement. However, a thorough review of the records showed that almost all of these patients arrived in a moribund condition and, after family discussions, were designated CMO. In 2004, 28 of the 31 patients were designated CMO, DNR/DNI or declared brain dead within a few days of admission. An additional two patients were in the process of being designated CMO when they died. The patient who died of the hemothorax after a central venous line placement had already triggered an institution-wide change in policy regarding central venous line placement that resulted in additional training and accreditation of physicians placing central lines.
Our analysis identified little that we could institute to improve the outcome in this cohort of patients. No apparent differences were found when comparing age, gender, percent of transfers, code status, diagnoses or complications. No change in attending neurosurgery staff or policy occurred during the 2003–2004 interval, but residents change services yearly at the institution. Perhaps an earlier initiation of family discussion would have contributed to shorter lengths of stay.
While the stated goal of pay-for-performance programs is to increase healthcare quality by applying specific measures and awarding better results, clearly this presently is not the case. Further, general availability of these results and rankings can adversely influence an institution’s public image and, in the current digital era, this information might negatively impact patient referrals or patient volume.
Conclusions
A beneficial role certainly exists in the healthcare field for quality measures
and standardization of some aspects of clinical care. However, these measures
should accurately and reliably reflect quality of care and adjust for factors
that inherently vary in the care of individual patients. Some medical fields
have quality indicators derived from admission or discharge coding that do
account for the severity of disease. Currently the Veterans Health Administration
tracks risk-adjusted mortality associated with neurosurgical procedures through
the Neurosurgery Consultants Board (8). Additional efforts should similarly
review and expand quality criteria for the field of neurosurgery.
Ideally, centralized data collection would provide the greatest accuracy when evaluating standard of care and would allow for the most precise risk stratification. Data on the number of cases transferred into and out of an institution also could prove useful as a quality indicator.
Lastly, reported results should be simplified into broader categories whether or not standards of care are met, and ranked lists perhaps should be avoided. In Tufts-NEMC’s review, the worse mortality rate from 2004 was not statistically significant as compared to the cohort. However, it is unlikely that consumers will consider statistical variability if ranked results are published. Fair reporting may be better achieved after a period of nonpublic quality evaluations to allow time for further corrective adjustments.
The opportunity currently exists for neurosurgeons to learn from the experience of other medical specialties and to direct the development of neurosurgical quality indicators. The involvement of neurosurgeons in the development of neurosurgical quality indicators is essential if these measures are truly to improve the quality of care for our patients.
REFERENCES
1. Agazzi S, de Tribolet N, Uske A, Regli L: Quality of life after aneurysmal subarachnoid hemorrhage in the elderly. Cerebrovasc Dis 18(2):174–175, 2004
2. Andrus CH, Villasenor EG, Kettelle JB, Roth R, Sweeney AM, Matolo NM: “To Err Is Human”: uniformly reporting medical errors and near misses, a naive, costly, and misdirected goal. J Am Col Surg 196(6):911–918, 2003
3. Bardach NS, Olson SJ, Elkins JS, Smith WS, Lawton MT, Johnston SC: Regionalization of treatment for subarachnoid hemorrhage: a cost-utility analysis. Circulation 109(18):2207–2212, 2004.
4. Birkmeyer JD, Dimick JB: Potential benefits of the new Leapfrog standards: effect of process and outcomes measures. Surgery 135(6):569–575, 2004
5. Birkmeyer JD, Finlayson EV, Birkmeyer CM: Volume standards for high-risk surgical procedures: potential benefits of the Leapfrog initiative. Surgery 130(3):415–422, 2001
6. Casson AG, van Lanschot JJ: Improving outcomes after esophagectomy: the impact of operative volume. J Surg Oncol 92(3):262–266, 2005
7. Cowan JA, Dimick JB, Henke PK, Huber TS, Stanley JC, Upchurch GR: Surgical treatment of intact thoracoabdominal aortic aneurysms in the United States: hospital and surgeon volume-related outcomes. J Vasc Surg 37(6):1169–1174, 2003
8. DePalma RG: Surgical Quality Programs in The Veterans Health Administration. General Surgery News 2006; November Supplement:27–31, 2006
9. Doran T, Fullwood C, Gravelle H, Reeves D, Kontopantelis E, Hiroeh U, Roland M: Pay-for-performance programs in family practices in the United Kingdom. N Engl J Med 355(4):375–384, 2006
10. Dudley RA, Johansen KL, Brand R, Rennie DJ, Milstein A: Selective referral to high-volume hospitals: estimating potentially avoidable deaths. JAMA 283(9):1159–1166, 2000
11. Duke BJ, Kindt GW, Breeze RE: Outcome after urgent surgery for grade IV subarachnoid hemorrhage. Surg Neurol 50(2):169–172, 1998
12. Finlayson SR. The volume-outcome debate revisited. General Surgery News 2006; November Supplement:64-68, 2006
13. Foerch C, Lang JM, Krause J, Raabe A, Sitzer M, Seifert V, Steimetz H, et al.: Functional impairment, disability, and quality of life outcome after decompressive hemicraniectomy in malignant middle cerebral artery infarction. J Neurosurg 101(2):248–254, 2004
14. Grover FL, Cleveland JC, Shroyer LW: Quality improvement in cardiac care. Arch Surg 137:28–36, 2002
15. Gupta R, Connolly ES, Mayer S, Elkind MS: Hemicraniectomy for massive middle cerebral artery territory infarction: a systematic review. Stroke 35(2):539–543, 2004
16. Havens DH, Boroughs L: “To Err is Human”: a report from the Institute of Medicine. J Pediatr Health Care 14(2):77–80, 2000.
17. Hunt WE, Hess RM: Surgical risk as related to time of intervention in the repair of intracranial aneurysms. J Neurosurg 28(1):14–20, 1968
18. Hutchinson PJ, Power DM, Tripathi P, Kirkpatrick PJ: Outcome from poor grade aneurysmal subarachnoid haemorrhage—which poor grade subarachnoid haemorrhage patients benefit from aneurysm clipping? Br J Neurosurg 14(2):105–109, 2000
19. Kassell NF, Torner JC, Haley EC, Jane JA, Adams HP, Kongable GL: The International Cooperative Study on the Timing of Aneurysm Surgery. Part 1: Overall management results. J Neurosurg 73(1):18–36, 1990
20. Kohn LT, Corrigan JM, Donaldson MS (eds): To Err Is Human: Building a
Safer Health System. Washington, D.C.: National
Academy Press, 2000
21. Malm J, Bergenheim AT, Enblad P, Hardemark HG, Koskinen LO, Naredi S, et al.: The Swedish Malignant Middle cerebral artery Infarction Study: long-term results from a prospective study of hemicraniectomy combined with standardized neurointensive care. Acta Neurol Scand 113(1):25–30, 2006
22. Mendelow AD, Gregson BA, Fernandes HM, Murray GD, Teasdale GM, Hope DT, et al.: Early surgery versus initial conservative treatment in patients with spontaneous supratentorial intracerebral haematomas in the International Surgical Trial in Intracerebral Haemorrhage (STICH): a randomised trial. Lancet 365(9457):387–397, 2005
23. Morgenstern LB, Frankowski RF, Shedden P, Pasteur W, Grotta JC: Surgical treatment for intracerebral hemorrhage (STICH): a single-center, randomized clinical trial. Neurology 51(5):1359–1363, 1998
24. Morrissey J: Patient safety proves elusive. Mod Healthc 34(44):6–7, 24–25, 28–32, 2004
25. Rabinstein AA, Mueller-Kronast N, Maramattom BV, Zazulia AR, Bamlet WR, Diringer MN, et al.: Factors predicting prognosis after decompressive hemicraniectomy for hemispheric infarction. Neurology 67(5):891–893, 2006
26. Robinson AR, Hohmann KB, Rifkin JI, Topp D, Gilroy CM, Pickard JA, et al.: Physician and public opinions on quality of health care and the problem of medical errors. Arch Intern Med 162(19):2186–2190, 2002
27. Safavi K: Unintended consequences: getting the most while avoiding the worst. J Healthc Manag 51(6):355–359, 2006
28. Stelfox HT, Palmisani S, Scurlock C, Orav EJ, Bates DW. The “To Err is Human” report and the patient safety literature. Qual Saf Health Care 15(3):174–178, 2006