Introduction

This concept describes four methods for defining long-term patients in acute care facilities. A statistical comparison of these approaches is also included.

Note: Long Term Care hospitals such as Deer Lodge and Riverview should be referred to as 'Long Term Care facilities' and should include a note (as a footnote or in the text) that they provide assessment and treatment, rehabilitation, chronic and personal care.

1. Length of Stay

Long-stay care or non-acute care inpatients have lengths of stay greater than 59 days. There is no distinction made between long-stay patients and long term care patients.

SAS code:

if los >= 60 then longstay = 1;

There are two concerns with using this definition of long-term care:


1. Short-stay non-acute hospital cases (e.g., some chronic care, respite care) cannot be detected. Furthermore, acute care with long lengths of stay will be misclassified as long-term care.
   
   
2. It may not be appropriate to consider total hospital stays given the fact that costs are calculated on an annual basis and hospital stays may exceed 365 days. A 365 day cut-off or use of in-year days should be considered.
   
   Notes :
   • A 45-day cutoff has also been used to define long hospital stays.
     
   • A chart-review study [ De Coster et al. (1996) ) by MCHP found that after 30 days, only 20% of medical patients still needed acute care; thus, several more recent reports have used a 30-day cutoff ( De Coster et al. (2000) )

2. Primary Service Classification

(from Tataryn et al., 1994)

The primary service classification variable (primpsvc) on the hospital claim record was used to define long-term inpatient hospital admissions to acute care facilities for mental health disorders. No consideration was given to the coding quality of the hospital. Each hospital admission was classified using the format described in the SAS Code example provided.


• MHMIS institutions included: ('0046') Selkirk Mental Hospital, ('0125') Eden Mental Health Centre in Winkler, and ('0047) Brandon Mental Health Centre.
  
  
• 'PSYCH. BED' was defined as the Primary Service Code designations for Psychiatry ('64') and Pediatric Psychiatry ('65').
  
  
• 'LONG TERM BED' was defined as any hospital claim where the primary diagnosis was an eligible mental health disorder (ICD-9-CM 290-319) and the first two digits of the primary service classification were '09' (Personal Care Unit) or '73' (Extended Treatment). All Mental Health stays at Deer Lodge and Municipal Hospital were defined as long-term stays using this definition.

3. Excluding Chronic and Rehabilitation Institutions

(from Roos and Shapiro, 1995)

Inpatient hospital care was classified as either short-stay inpatient care or longstay care depending on the number of days spent in hospital. Hospital stays less than 60 days were defined as short-stay. Long-stay was defined as 60 days or longer.

Note: This definition differs from length of stay (LOS) in one major way: chronic and rehabilitation institutions were not included in the analyses when reference was made to "longstay patients". Thus, ('0006') Municipal (Riverview) hospital, ('0019') Deer Lodge Centre, ('0017') Rehabilitation Centre for Children and the ('0020') Adolescent Treatment Centre were excluded from the analyses as they are considered non-acute facilities (i.e. they do not admit acute-care patients).


• The term 'acute care' is used interchangeably with the term short stay inpatient care in the report.

St. Boniface ('0005') and Seven Oaks ('0011') Hospitals have geriatric units for longer-term Rehabilitation and thus non-acute stays can be identified by the Service Code. Also, the Rehabilitation hospital of HSC ('0016') has a large number of trauma patients for whom Rehab takes a long time and thus should not be considered acute care. Roos and Shapiro (1995) counted all the patients at St. Boniface, Seven Oaks, and the Rehabilitation at HSC as acute.

4. Hospital Service Codes

(from Shanahan et al., 1994 )

Only Acute Care hospitals were considered. The following facilities were excluded (in addition to the "poor coding" hospitals): ('0006') Winnipeg Municipal, ('0019') Deer Lodge, ('0017') Rehab for Children, ('0020') Manitoba Adolescent Treatment Centre, ('0117') Cartwright District, ('0127') Elkhorn Medical Nursing Unit, ('0142') Hartney Medical Nursing Unit.

SAS code:

/** Select Hospitals, and remove long term care hospitals **/
if hosp < =212 and hosp in (108 127 142 19 6 17 20);

Hospital service codes were used to count the number of cases classified as non-acute. The following service codes were used:

Service codes:	Sub-Service codes:
'09' = personal care unit '72' = geriatrics '73' = long term care '70' = physical medicine and rehab '71' = pediatric physical medicine and rehab	'93' = paneled for chronic '94' = rehab '95' = social '96' = assessment '97' = chronic '98' = respite '99' = paneled for pch

These service codes were used based on results from a survey of Manitoba hospitals conducted by Marni Brownell and Julene Reimer at Manitoba Health. The survey findings and recommendations were used to identify the hospitals that were good coders (i.e. long-term care patients were easily identifiable). Specifically, good coder hospitals were designated as such if the days counted using their service codes compared with data from Manitoba Health revenue days (extended care, respite and paneled from HS1 forms). If the match was 70% or better the service codes were used for that hospital.

The seven Service Code fields (1 primary and 6 sub-service classifications) in the Hospital Discharge Abstracts data were examined for the service codes listed above. If any one of these codes were used in any of the fields the patient was classified as a non-acute case and the days associated with the non-acute code(s) were recorded.

Specific service codes were also used in 3 hospitals:



1. Health Sciences Centre:

   (hosp=16 and ps(i)) in
   '3484' (Orthopedic surgery - RehablResp Centre)
   '1894' (Respirology - RehablResp Centre)
   '5918' (Oncology - RehablResp Centre)
   (hosp=16 and substr(ps(i),1,2) in
   '78' (Spinal Injury - RehablResp Centre)
   '79' (Amputation - RehablResp Centre)
   '80' (Stroke -RehablResp Centre)
   '81' (Brain Injury - RehablResp Centre)
   '82' (Neuromuscular - RehablResp Centre)
   '83' (Musculoskeletal - RehablResp Centre)
   '85' (Rheumatology - RehablRespCentre)
   '86' (Tuberculosis - RehablResp Centre)
   '87' (Chronic Obstructive Lung Disease RehablResp Centre)

   These codes indicate some type of extended care at HSC, but not all urban hospitals.
   
   

2. Seven Oaks:

   (hosp = 0011 and substr (ps(I),1,2) = 84) as reported by the hospital.

3. St. Boniface:

   (hosp = 0005 and substr (ps(I), 1,2) ='65') (Pediatric Psychiatry)

   Note:
   
   Incorrect Service Code allocation by "poor coding" hospitals may occur for the following reasons:
   
   

   • the paneled or long-stay patient may have been charged for the hospital stay, and their physician wanted to save them money, so the inappropriate Service Code was used.
     
   • incomplete documentation by the physician
     
   • poor hospital coding practices, or
     
   • some other reason.
   
   Limitation:
   
   Thirty-two of the provincial hospitals (all of them rural) do not accurately code long term care and paneled care on the Hospital Service Abstract.
   
   Two possible solutions:
   
   
   1. Continue the practice of using long-term care days reported by Manitoba Health. This would prevent comparisons of rates built on individual claims (e.g., surgical/medical rates) - i.e., it provides hospital level data only.
      
      
   2. It may be possible to use the case mix for the "good coding" hospitals to assign long term days to cases in "poor coding" hospitals. However, the results do not produce completely satisfactory results.
   
   

   Three problems encountered using case mix:
   
   

   1. The total number of claims for long-term care for the "poor coders" cannot be counted.
      
   2. Some of the "poor coders" may not have corresponding RDRGs®. A solution to this problem was attempted by using only the proportion of the RDRGs® that occur in the "poor coding" hospitals.
      
   3. More long-term days than is appropriate may be mistakenly assigned to a claim. There are between 0% and 10% of the claims that receive more days from the long-term care than the actual total length of stay. A possible solution involves taking these extra days and re-distributing them to the remaining RDRGs®.

   Note: The solution proposed above depends on accurate data being provided by Manitoba Health.

Other Issues

1. The FIS (Financial Information System) and HS1 methods are not reliable means of obtaining chronic care information. A new sub-service code (patient paneled for chronic denoted by '93') was added for the Winnipeg Region on April 1, 1993.
   
   
2. Patient Service Code '73' has been changed to 'Long Term Care' from 'Extended Treatment'. The code continues to be used for patients in designated Long Term Care beds as was done previously. It has been recommended that the following Service Codes and Sub-Service Codes be classified as "nonacute":

   Service Codes	Sub-Service Codes
   '09' = personal care unit '72' = geriatrics '73' = long term care '70' = physical medicine and rehab	'93' = paneled for chronic '94' = rehab '95' = social '96' = assessment '97' = chronic '98' = respite '1991' = paneled for pch

Methodology Comparisons

The methods described above* were 'tested' on the total number of long-term care claims (separations) in good coder acute hospitals in the province during 1991-92 using:

Note: * The chronic and rehabilitation definition was not included because it was too similar to the length of stay definition.

Hospital Discharge Abstract data from 1991/92 and 1992/93 were used to capture the claims made in 1991/92.



• During 1991/92 there were 148,668 acute hospital inpatient separations in Manitoba.
  
• The length of stay definition captured 4,241 cases of non-acute care, including 1,953 unique cases (i.e. the other definitions did not identify them).
  
• The hospital service code definition captured 6,528 cases of non-acute care, including 2,903 unique cases.
  
• The primary service classification definition captured 1,784 cases. None were unique.
  
• Only 447 cases were identified using all three definitions.

Cross tabs for 91/92 for Long Term Care Definitions



Costing Method	Length of Stay < 60	Primary Service Class	Case Count	Percentage	Cumulative Case Count
0	0	0	140,207	94.3	140,207
0	1	0	1,953	1.3	142,160
1	0	0	2,903	2.0	145,063
1	0	1	1,337	0.9	146,400
1	1	0	1,841	1.2	148,241
1	1	1	447	0.3	148,688