PEDIATRICS Vol. 105 No. 4 April 2000, pp. 733-737

Pediatric Milliman and Robertson Length-of-Stay Criteria: Are They Realistic?

Marion R. Sills, MD, MPH^*, , Zhihuan J. Huang, MD, MPH^*, Cheng Shao, MD, MPH^*, Mark F. Guagliardo, PhD^*, James M. Chamberlain, MD^*, and Jill G. Joseph, MD, PhD^*

From ^* Children's National Medical Center, Washington, DC; and  Children's Hospital, Denver, Colorado.

	ABSTRACT

Top Abstract Methods Results Discussion References

Objective.  Guidelines for inpatient length of stay (LOS) have been developed by several groups; among the most widely applied are those published by Milliman and Robertson (M&R). Few published reports have examined the relationship of actual practice to such guidelines, none in pediatric populations. This study was designed to compare pediatric practice in a large and defined population to M&R LOS criteria.

Methods.  Administrative data from New York State in 1995 were used to examine LOS for discharges corresponding to 16 selected pediatric diagnoses for which M&R publishes guidelines. Outliers, defined as the 2% of discharges with the longest LOS, were eliminated. The distribution of LOS for each diagnosis was compared with M&R LOS guidelines.

Results.  In New York State during 1995, pediatric LOS was markedly divergent from M&R guidelines. In general, the percentage of discharges in excess of the criterion LOS was less for nonmandatory admissions (croup: 23%, gastroenteritis: 44%, and pneumonia: 48%) than for those requiring surgery (uncomplicated appendectomy: 67%, pyloromyotomy: 62%, and major but noncritical burns: 64%) or prolonged treatment with antibiotics (bacterial meningitis: 91% and osteomyelitis: 86%).

Conclusions.  In New York State during 1995, LOS for selected pediatric conditions was generally in excess of published M&R guidelines. This raises concern about the potential effects of such guidelines on both patients and the hospitals caring for them. While endorsing the need for cost-effective practice, we call attention to the methods used to develop and validate guidelines.length of stay, pediatrics, managed health care, administrative data, practice guidelines.

Systematic clinical cost-containment efforts have prompted the rapid development and dissemination of clinical practice guidelines nationwide.^1-7 Although practice guidelines regarding appropriate length of stay (LOS) have been widely promulgated, there are limited data addressing patient outcomes after their implementation. Little is known regarding their impact on health care costs. For example, do they cost-shift to postdischarge care, resulting in greater net costs?⁸ How do they impact the financial viability of health-care facilities? More important, however, are questions about their impact on patient health and satisfaction? Several studies have demonstrated that certain LOS-related guidelines do not adversely impact patient health outcomes.^9-20 Other studies and numerous commentaries in both the lay and medical press have raised concerns regarding the largely unknown impact of guidelines on the health of more vulnerable populations, such as the elderly, young, or chronically ill.^21-26 Despite the absence of careful outcome evaluation, the implementation of LOS guidelines and other cost-containment efforts goes largely unmeasured. Furthermore, many LOS guidelines are not rigorously evidence-based, often drawn more from benchmarking comparisons between institutions or group consensus than from epidemiologic data.

One commonly referenced guideline for pediatric LOS was developed by the actuarial firm of Milliman and Robertson (M&R). The M&R guidelines are used, in concert with others, by most major managed care organizations, including Group Health Cooperative, Harvard Pilgrim Health Care, Kaiser Foundation Health Plan, and Pacificare. Faculty members in the Department of Pediatrics at the University of Texas-Houston Medical School developed the M&R benchmarks for expected pediatric LOS with the intention of focusing on best practice rather than on median performance. The guidelines for acute care are based on treating routine, uncomplicated cases,²⁷ although informal and unpublished reports abound of their application to all cases without regard to medical or psychosocial complexities (J. Gibson, personal communication, 1999).

The purpose of this investigation was to compare M&R pediatric LOS guidelines to actual LOS among pediatric patients with specific but diverse diagnoses.

	METHODS

Top Abstract Methods Results Discussion References

To quantify pediatric inpatient LOS, data were abstracted from the 1995 New York Statewide Planning and Research Cooperative System (SPARCS) database. This administrative database provides comprehensive, uniform, case-based data for each inpatient hospitalization in a nonmilitary, acute-care hospital in the state. The study population for this article consists of patients from infancy (<1 year old) through 17 years old. We eliminated newborn admissions (ie, admissions starting at the time of birth) because of their unique set of diagnoses and their correspondingly unusual LOS distribution. LOS is uniformly defined in the SPARCS database on a calendar-day basis, so that a patient admitted in the evening and discharged the next morning would have a LOS of 2 days.

M&R provide LOS guidelines for 59 pediatric inpatient diagnoses, defined by International Classification of Diseases, 9th Revision clinical modification codes.²⁸ For each diagnosis, both a functional LOS and goal LOS are identified. The functional LOS consists of a day-by-day idealized schedule for clinical care during the admission. The goal LOS represents a target for the hospital and is always 1 day shorter than the functional LOS for each diagnosis. For the purpose of this analysis, we chose the longer, functional LOS recommendations that are more related to patient care than the shorter goal LOS guidelines, which are more pertinent to compensation for care.

We assessed the frequency distribution of all 59 M&R diagnoses among pediatric discharges in the SPARCS database. From the 45 of these diagnoses containing at least 100 discharges during 1995, we selected a sample of 16 diagnoses based on criteria designed to provide diversity. First, we sought to achieve a mix of higher (major burn, complicated appendicitis, and bacterial meningitis) and lower (croup, cellulitis, and gastroenteritis) severity diagnoses. Our second criterion was to include diagnoses specifically relevant to infants (croup, pyloromyotomy, and bronchiolitis) as well as to the general pediatric population. Third, we sought to include both medical (cellulitis, osteomyelitis, pneumonia, sickle cell disease, asthma, bronchiolitis, croup, diabetic ketoacidosis, gastroenteritis, meningitis, and pyelonephritis) and surgical (appendicitis, pyloromyotomy, and major burn) conditions, with at least 1 of each requiring long-term antibiotic therapy (bacterial meningitis, osteomyelitis, and complicated appendicitis).

Because M&R guidelines include consideration of the patient requiring intensive care (eg, for the patient with diabetic ketoacidosis), we did not remove patients on these grounds alone. Nonetheless, we wanted to minimize the effects of extreme outliers on the distributions of LOS, while nonetheless preserving information about the range of LOS in actual pediatric practice. Selecting an appropriate threshold for exclusion of discharges with extreme LOS was further complicated because we were considering multiple diagnoses, some with limited numbers of discharges, and we wanted to apply a reasonable but uniform cutoff. Although various thresholds might be reasonable, after inspection of the distributions and consideration of the literature, we chose to exclude the 2% of discharges with longest LOS. The choice was somewhat arbitrary but has been applied by others in the analysis of administrative data.²⁹

Using SAS (SAS Institute, Cary, NC), we examined the mean, median, mode, and range of LOS for each of the 16 diagnoses. For each diagnosis, we also calculated: 1) the percentage of discharges with LOS longer than the M&R functional LOS guideline, and 2) the percentage of hospital-days that exceeded the M&R functional LOS guideline.

	RESULTS

Top Abstract Methods Results Discussion References

As shown in Table 1, the number of available cases for analysis of the 16 selected diagnoses varied from 22 412 for asthma to 138 for bacterial meningitis. Although the range in LOS is large, for 10 of the conditions the modal LOS corresponds to or is less than the M&R functional LOS guideline. In contrast, the median exceeds the guideline LOS for all but 3 common pediatric conditions, none of which require surgery or prolonged antibiotics: croup, gastroenteritis, and pneumonia. As would be expected for highly skewed LOS data, the mean LOS generally exceeds the median, often dramatically so, as in the case of bacterial meningitis and osteomyelitis. To describe better the distribution of LOS, curves were constructed for each of the 16 diagnoses with the M&R LOS guideline imposed on each as a vertical line, as shown in Fig 1. The shape of specific LOS distributions often seems related to the standard practice. For example, the modal LOS for bacterial meningitis is between 9 and 10 days, while for complicated appendicitis it is 7 days. Distributions are expectedly more stable for those diagnoses with a larger number of discharges, such as asthma, than for those with fewer, such as osteomyelitis.

View larger version (25K): [in this window] [in a new window]   Fig. 1.   Distribution of LOS for selected pediatric diagnoses, New York State, 1995.

To examine further the relationship of practice to the published guidelines, we calculated the proportion of discharges and of bed-days in excess of the functional LOS criteria. The use of bed-days represents the cumulative burden on institutions caring for patients if M&R criteria were being used as the basis for reimbursement. That is, whether 5 patients each stay 2 days beyond the guidelines or 2 patients stay 5 days beyond, the admitting institution would be responsible for 10 days of uncompensated care. Shown in Fig 2 are the percentage of patients and bed-days for each diagnosis that exceeded the M&R LOS guidelines. In every instance, the excess in bed-days is greater than in individual discharges, reflecting the effects of those patients with longer LOS. In reviewing these data, it is important to recall that we used the functional LOS criteria for the 16 diagnoses rather than the goal LOS, which is shorter by 1 day in every instance. Thus, these analyses provide a conservative measure of the discrepancy between guidelines and practice.

View larger version (34K): [in this window] [in a new window]   Fig. 2.   Percentage of patient discharges and bed-days in New York State, 1995 exceeding M&R LOS guidelines.

	DISCUSSION

Top Abstract Methods Results Discussion References

Our findings demonstrate that actual pediatric LOS in New York State during 1995 exceeded, often markedly, the M&R functional LOS guidelines. The discrepancy between utilization data from SPARCS and M&R guidelines was most marked in diagnoses requiring long courses of antibiotics (bacterial meningitis, osteomyelitis, and complicated appendectomy) as well as in surgical diagnoses (complicated appendectomy, major burns, and pyloromyotomy). M&R LOS criteria are based on the recommendation for early discharge of patients with high quality and sophisticated home care arrangements, notably including home administration of intravenous antibiotics. Inconsistent availability or referral to such posthospital care may explain much of this difference. In contrast, minimal differences between the guidelines were noted for 3 common pediatric conditions, none of which is considered mandatory³⁰ and each of which can be treated by general pediatricians without referral to subspecialist care.

The wide discrepancy between the M&R guidelines and actual practice is consistent with other reports. A study of 25 surgical diagnoses revealed similar findings: in 8 of 25 diagnostic groups, the observed mean LOS exceeded M&R guidelines by >5 days.³¹ Similarly, a review of the 85 728 patients in the National Trauma Data Bank found that the M&R LOS recommendation for each diagnostic category "grossly underestimates the LOS observed in patients" and that "patients in the National Trauma Data Bank who actually met M&R guidelines had a significantly higher mortality" than did those whose LOS exceeded the recommendation.³²

These discrepancies between M&R guidelines and observed LOS data raise important and provocative questions regarding the evolution of guidelines and of collaborative practices more generally. First, what is the role of data in developing guidelines? Like any change in medical practice, guidelines are best when developed and evaluated in peer-reviewed clinical trials. For example, evidence-based and prospectively validated guidelines have been developed for adults hospitalized with community-acquired pneumonia³³ and upper gastrointestinal tract hemorrhage.¹² Both studies provide clear documentation that allow clinicians to critically appraise how they were developed, implemented, and evaluated.

In contrast, accumulation of knowledge through traditional, randomized clinical trial methods is slow and responds poorly to rapid changes in the health care system. Therefore, the task is to balance potentially competing demands for speedy and reasonable change. Both the Institute of Medicine³⁴ and the Agency for Health Care Policy and Research³⁵ have set high standards for the development of guidelines, including the involvement of multidisciplinary panels and the use of an explicit, evidence-based approach. This is the methodology used by governmental groups such as Institute of Medicine and Agency for Health Care Policy and Research as well as by the American Academy of Pediatrics. Should the same be expected of consulting firms that develop and market clinical guidelines? At a minimum, should we expect that the data and methods contributing to guideline development be available for public discussion and debate? If so, how are the proprietary interests of commercial ventures to be protected?

A second set of questions concerns the effects of LOS guidelines on both cost and quality of care. Does LOS shorten over time as LOS guidelines penetrate a market? If so, which diagnoses respond most quickly and which respond slowly or not at all? What barriers or clinical practices can be modified and which do not change in the face of LOS guidelines? Can guidelines designed to reduce cost also improve quality? Such questions require examination of longitudinal data. We note, however, that the M&R guidelines specify well-defined home health care requirements for diagnoses postdischarge that may not be realistic. For example, in conditions such as bacterial meningitis, significant barriers to the recommended discharge on hospital-day 3 may exist. There may be a limited number of home health care agencies that would accept responsibility for assessing hemodynamic stability for a child discharged with bacterial meningitis, as suggested in the M&R criteria.³⁶ These issues related to the feasibility of the guidelines merit further careful study.

Third, how easily can guidelines be implemented in settings different from that in which they were developed? The M&R guidelines were developed at a single health care organization, University of Texas-Houston Medical School, and may not be applicable in the hospital settings and pediatric populations of New York State. Similarly, we recognize that analyses reported here based on New York State data may not apply to the rest of the country. Indeed, Milliman and Robertson's state-by-state comparisons of LOS efficiency find New York State the most extreme example of inefficiency, with LOS 88% higher than the country's most efficient states.³⁷ It is apparent that similar analyses are needed of other state or national databases.

LOS guidelines may help clinicians optimize care, reduce costs, and potentially improve patient satisfaction. However, both the public and policymakers are increasingly concerned that the recent pressures for cost-savings have come to dominate the practice of medicine. As but 1 example, the issue of drive-by mastectomies generated considerable attention and some legislative changes. Nonetheless, few data are available to inform this discussion and even fewer are relevant to the large group of vulnerable pediatric patients. Although we enthusiastically endorse the notion of increasing the efficiency of inpatient care, we are concerned that any process of goal-setting be informed by available data, that it be as open as possible, and that any effects on patients and families be carefully examined.

	ACKNOWLEDGMENTS

Support for this research was provided by the Child Health Accountability Initiative and by the Children's Research Institute of Children's National Medical Center, Washington, DC.

	FOOTNOTES

Received for publication Nov 18, 1999; accepted Jan 7, 2000.

Reprint requests to (J.G.J.) Children's National Medical Center, 111 Michigan Ave NW, Suite 5500, Washington, DC. E-mail: jjoseph{at}cnmc.org

	ABBREVIATIONS

LOS, length of stay; M&R, Milliman and Robertson; SPARCS, New York Statewide Planning and Research Cooperative System.

	REFERENCES

Top Abstract Methods Results Discussion References

1. Brook RH Practice guidelines and practicing medicine: are they compatible? JAMA 1989; 262:3027-3030 [Abstract]
2. Woolf SH Practice guidelines: a new reality in medicine. I. Recent developments. Arch Intern Med 1990; 150:1811-1818 [Abstract]
3. Audet AM, Greenfield S, Field M Medical practice guidelines: current activities and future directions. Ann Intern Med 1990; 113:709-714
4. Pearson SD, Goulart-Fisher D, Lee TH Critical pathways as a strategy for improving care: problems and potential. Ann Intern Med 1995; 123:941-948 [Abstract/Free Full Text]
5. Yandell B Critical paths at Alliant Health System. Qual Manage Health Care 1995; 3:55-64
6. Falconer JA, Roth EJ, Sutin JA, The critical path method in stroke rehabilitation: lessons from an experiment in cost containment and outcome improvement. Qual Rev Bull 1993; 19:8-16
7. Grimshaw JM, Russell IT Effect of clinical guidelines on medical practice: a systematic review of rigorous evaluations. Lancet 1993; 342:1317-1322 [CrossRef][Medline]
8. Ellrodt AG, Conner L, Riedinger M, Weingarten S Measuring and improving physician compliance with clinical practice guidelines: a controlled interventional trial. Ann Intern Med 1995; 122:277-282 [Abstract/Free Full Text]
9. Ray WA, Griffin MR, Baugh DK Mortality following hip fracture before and after implementation of the prospective payment system. Arch Intern Med 1990; 150:2109-2114 [Abstract]
10. Gerety MB, Soderholm-Difatte V, Winograd CH Impact of prospective payment and discharge location on the outcome of hip fracture. J Gen Intern Med 1989; 4:388-391 [Medline]
11. Weingarten S, Riedinger MS, Sandhu M, Can practice guidelines safely reduce hospital length of stay? Results from a multicenter interventional study. Am J Med 1998; 105:33-40 [CrossRef][Medline]
12. Hay JA, Maldonado L, Weingarten SR, Ellrodt AG Prospective evaluation of a clinical guideline recommending hospital length of stay in upper gastrointestinal tract hemorrhage. JAMA 1997; 278:2151-2156 [Abstract]
13. Hay JA, Lyubashevsky E, Elashoff J, Maldonado L, Weingarten SR, Ellrodt AG Upper gastrointestinal hemorrhage clinical-guideline determining the optimal hospital length of stay. Am J Med 1996; 100:313-322 [CrossRef][Medline]
14. Weingarten SR, Conner L, Riedinger M, Alter A, Brien W, Ellrodt AG Total knee replacement: a guideline to reduce postoperative length of stay. West J Med 1995; 163:26-30 [Medline]
15. Weingarten SR, Riedinger MS, Conner L, Practice guidelines and reminders to reduce duration of hospital stay for patients with chest pain: an interventional trial. Ann Intern Med 1994; 120:257-263 [Abstract/Free Full Text]
16. Kong GK, Belman MJ, Weingarten S Reducing length of stay for patients hospitalized with exacerbation of COPD by using a practice guideline. Chest 1997; 111:89-94 [Abstract/Free Full Text]
17. Weingarten S, Riedinger M, Conner L, Johnson B, Ellrodt AG Reducing lengths of stay in the coronary care unit with a practice guideline for patients with congestive heart failure: insights from a controlled clinical trial. Med Care 1994; 32:1232-1243 [Medline]
18. Weingarten S, Riedinger M, Conner L, Hip replacement and hip hemiarthroplasty surgery: potential opportunities to shorten lengths of hospital stay. Am J Med 1994; 97:208-213 [CrossRef][Medline]
19. Weingarten S, Agocs L, Tankel N, Sheng A, Ellrodt AG Reducing lengths of stay for patients hospitalized with chest pain using medical practice guidelines and opinion leaders. Am J Cardiol 1993; 71:259-262 [CrossRef][Medline]
20. Rhew DC, Riedinger MS, Sandhu M, Bowers C, Greengold N, Weingarten SR A prospective, multicenter study of a pneumonia practice guideline. Chest 1998; 114:115-119 [Abstract/Free Full Text]
21. Kassirer JP The quality of care and quality of measuring it. N Engl J Med 1993; 329:1263-1265 [Free Full Text]
22. Relman AS Reforming the health care system. N Engl J Med 1991; 324:920-922 [Medline]
23. Wilson MC, Tunis SR, Hayward RS, Kern DE, Howard DM, Bass EB Primary care physicians' attitudes toward clinical practice guidelines. Med Decis Making 1991; 11:334
24. Fitzgerald JF, Moore PS, Dittus RS The care of elderly patients with hip fracture: changes since implementation of the prospective payment system. N Engl J Med 1988; 319:1392-1397 [Abstract]
25. Fitzgerald JF, Fagan LF, Tierney WM, Dittus RS Changing patterns of hip fracture care before and after implementation of the prospective payment system. JAMA 1987; 258:218-221 [Abstract]
26. Mitchell WH Who are Milliman & Robertson and how did they get in my face? J Ky Med Assoc 1996; 94:521-522 [Medline]
27. Milliman and Robertson, Inc. Schibanoff JM, ed. Pediatric HSIM (health status improvement and management): caring for a community of children. Available at: http://www.milliman-hmg.com/publications/hmg/hmgordn1.html. Accessed: October 20, 1999
28. World Health Organization. International Classification of Diseases, 9th Revision. Geneva, Switzerland: World Health Organization; 1999
29. Markowitz JS, Pashko S, Gutterman EM, Death rates among patients hospitalized with community-acquired pneumonia: a reexamination with data from three states. Am J Public Health 1996; 86:1152-1154 [Medline]
30. McConnochie KM, Roghmann KJ, Liptak GS Socioeconomic variation in discretionary and mandatory hospitalization of infants: an ecologic analysis. Pediatrics 1997; 99:774-784 [Abstract/Free Full Text]
31. Rutledge R An analysis of 25 Milliman & Robertson guidelines for surgery: data-driven versus consensus-derived clinical practice guidelines. Ann Surg 1998; 228:579-587 [CrossRef][Medline]
32. Meredith JW, Burney R, Burton S, Milliman & Robertson length of stay guidelines are not appropriate for trauma patients: a comparison with the NTDB. J Trauma 1999; 47:208
33. Halm, EA, Fine MJ, Marrie TJ, et al Time to clinical stability in patients hospitalized with community-acquired pneumonia: implications for practice guidelines. JAMA 1998; 279:1452-1457 [Abstract/Free Full Text]
34. Eddy DM. Clinical decision making: from theory to practice. Guidelines for policy statements: the explicit approach. JAMA. 1990;263:2239-2240, 2243
35. Berg AO, Atkins D, Tierney W Clinical practice guidelines in practice and education. J Gen Intern Med 1997; 12:S25-S33
36. Schibanoff JM, ed. Pediatric Health Status Improvement and Management. Philadelphia, PA: Milliman and Robertson, Inc; 1998
37. Anonymous Use these targets to benchmark state, regional LOS efficiency. Data Strategy Benchmarks 1997; 1:43-44