Background

Identifying individual (solo) practices and group practices has come up regularly as part of MCHP research projects. When measuring continuity of care this is an important issue because a clinic may be considered a single point of care even though there may be several physicians providing service. Unfortunately there is no accepted or easy solution and there will likely be an ongoing need to identify patients going to individual physicians, clinics, or multiple physicians for primary care. Identifying group practices (clinics and physicians working together) is difficult from administrative data. There have been several attempts to look at this issue and this document provides an overview of the issues and possible starting points for further work.

Electronic site numbers or tape exchange codes (TC) have been available on medical billing claims since 1994 and might be used for identifying the practice location (see Use of Electronic User Site Locator data (EUSL) for identifying group practices (P. Nicol, source: Manitoba Health, 2004). The ESUL contains several tables which identify the relationship between TC and TC name, physician billing number, and base number. The EUSL also contains some doctor specific information similar to what is in the Provider Registry. The TC is an electronic billing number used for filing billing claims by physicians. These numbers may be associated with individual physicians, clinics, groups of physicians, and billing services. TCs (if it exists) may be used to associate a physician with a billing location.

Methods

1. Physician Resource Database (PRDB)

Based on PRDB documentation and work partially completed as part of Profiling Primary Care Practice in Manitoba (Frohlich et. al., 2006) . Note: groups were not used in the final deliverable but the development was explored.

See discussion in the MEDICAL GROUP CHARACTERISTICS (internal access only) portion of the PRDB (Bogdanovic, 2004) documentation. The data associated with this resource is available for 2000/01, 2001/02. If a research project requires more recent information the data will be updated at that time.

The medical group characteristics file in the PRDB provides some general information on each TC and the number of associated general practitioners [now referred to as family physicians].

An algorithm was developed to compute a weighted number of physicians based on the relative contribution of each physician by claim within the individual TC. The data also contains a variety of other measures such as the proportion of GPs, weighted FTE, workload within the TC, how long/often the TC was used, etc... The weighted numbers and other associated fields can be used by researchers to determine if the TC represents a group practice.

Steps used in building the resource and associated weights are outlined in the PRDB documentation. The medical billing data alone was used to identify all of the physicians claiming under the same TC. The use of the weighted TC information may be a superior method since it does not rely on identification of specific services or types of TC. Individual TCs were not characterized with regard to the type or service that was being provided. The use of the provided weights along with some manual identification (see below) may be the most appropriate route to identifying group practices in the future.

Based on a memo from N. Dik Feb 13, 2007. Note: groups were not used in the final deliverable but the development was explored.

Medical claims were selected for a single year (0102) with the following exclusions: Out of Province billings and residents, unknown or missing PHINs, non-GPs, North (Burntwood, NOR-MAN, Churchill) and Brandon areas were not part of this project. EUSL information including TC was joined to medical claims data by Base# or PHYNO*. TCs were reviewed manually to see if any additional exclusions were necessary (e.g. not a family practice). Those TCs were excluded from further consideration. TCs associated with Billing Services were identified: we first sought to re-assign all of them back to zero ('000'), but decided NOT to do that after some investigation (checked for % shared patients within each BS): only BS #402 (SWEP Management) was re-assigned (see minutes June 23 2004 and e-mail from Alan Katz (Oct 19 2004). The identification of solo/group practices was not completely resolved in this project.

The following steps were taken:

1. Select list of PHYNO that bill with and without TC. Some of them bill through > 1 TC.
   
   
2. Possibly adjust for Timing issues (see below under issues)
   
   
3. Exclude PHYNOs w/o Birth Date, that do not have associated TC. There were 10 of these in the data period used with this study.
   
   
4. Remaining claims after most basic exclusions (submitted by PHYNOs with PHISGRP**=1 with or without TC) were used to define practices, by summing up all claims by TC (if present) or by PHYNO (if no TC present).
   
   
5. Make some additional exclusions, based on Practice type, size or location (< 1000 claims per practice, located in North or in Brandon ). Exclude TCs that were non-primary care practices (16 in this study), as well as Northern Medical Unit ( TC list #1 ). This list of TCs will need to be reviewed and updated for other years of data.
   
   
6. If necessary, re-assign Billing Services practices to be TC=0 (no TC) PHYNOs (checked for % shared patients within each first). ( TC list #2 )
   
   
7. Exclude 1 PHYNO w/o Birth Date that bills through 24 different TCs (it will result in reduction in size for several practices associated with that PHYNO, as well as in removal of at least 10 Micro-Management Systems ( MMS )TCs.
   
   
8. Exclude 6 community clinics corresponding PHYNOs without Birth Date ( all located in North )
   
   
9. Include 10 clinics with TC that are also identified by PHYNO w/o Birth date (these were all salaried community clinics).

   There were 2 lists created:
   
   

   • With TC (Solo and Group practices)
   • Without TC (all considered Solo for now).

   
   TCs with 1 PHYNO per TC were considered Solo practices (with exception of 10 salaried Community Clinics). TCs with > 1 PHYNO per TC were considered Group Practices (will calculate % shared, workload FTE, visits and N of Patients for each). List of no-TC PHYNOs -- were all Solo practices, flagged as no-TC, size=1 PHYNO

10. Additional proposed exclusions for both TC practices and no-TC practices:
    
    • Those with < 1000 claims a year.
      
    • North/Brandon practices (that were not excluded by previous criteria).
      
    • TCs with total number of days worked < 100 (~5 sites) (Nov 19 meeting).
      
    • Look at GPs with >1 billing number. Some of them have low volume or > RHA during the year.
      
    • Claims from nursing stations? (N = 20 PHYNOs, 6 with low volume).
      
    • Apply exclusions by PHYNO to single practices (see Exclusion criteria by PHYNO in Exclusions.doc.) GPs that changed TEXTCODE (possible move?) - to be worked out.

Analyses Undertaken

1. To determine Group/Solo Practice.
   To determine if it was Group or Solo practice we looked at # of GPs billing through same TC. Out of initial >1000 GPs, ~ 400 were classified as Family Practices (~ 300 Solo + 100 Groups). Note: We could flag docs if they practice in a group in addition to individual practice.
   
   
2. % Shared patients
   Among all of GPs in a practice or pairs of GPs (see docs in folder Natalia/New Material for Nov 7 Meeting). We had 2 types of sharing calculated (based on 1 year of claims, but ideally should be at least 2 years, since many people do not go to the doctor every year):
   
   • Version One - where we counted patients with at least 1 claim (visit) and shared are counted if sharing occurred with at least 2 different doctors within TEXTCODE, independent of date. We thought to do a paired analyses, but found it too complicated. Not sure how to summarize the results back to a TC with 19 Docs and 171 possible pairs combinations (there were total 2109 pairs in selected 111 SITES (TC)).
     
   • Version Two - only patients who had at least 2 claims within selected TC were used - these patients have a potential to be shared. Again, we used no restriction by date.
   
   
   Comparing 2 above-mentioned numbers we noticed that for some the second % was much higher. We thought those might be Walk-Ins and Emergency, because basically very few patients go there more than once, and if they do, they see whoever is on call there, without appointments - more likely to be sharable.
   
   • Note 1 (Sharing). The possibility of counting shared excluding cases where MDs were sought as working successively (one MD retired and another took over) was discussed but not implemented.
     
   • Note 2 (Sharing). Sharing between practices was discussed but not implemented
   
   Abbreviations:
   
   
   • PHYNO - Unique Physician Number from PRDB. In the primary care physician group this is the base number from the Provider Registry data.
     
     
   • PHISGRP=1 - Defined in the PRDB. This identifies PHYNOs that are practicing under a general practice bloc of practice (11_).

3. Continuity of Care

Based on a memo from M. Sirski Feb 25, 2003.

(Menec et. al., 2005) looked at a specific number of clinics as well as solo practices for patients living in Winnipeg (9899b, 9900b). Individual physicians were identified by finding unique combinations of individual billing numbers and TC. Clinics were identified using specific TCs: TRANSCONA MEDICAL CLINIC (099), MEADOWOOD MEDICAL CENTRE (050), TUXEDO FAMILY MEDICAL CTR (263), MCPHILLIPS MEDICAL GROUP (044), CHARLESWOOD MED CLINIC (085), ST BONIFACE CLINIC (007) - 6 in total. See COC project NewClinic.doc - February 25, 2003.

It should be noted that this project used TC (identified clinics) as the identifier when measuring continuity of care. This is different than most other studies which use PHYNO. This was possible because a select number of group practices were identified instead of trying to identify group practices generally.

Steps Taken:


1. Input all ambulatory visits (ambvis='1') to GPs and pediatricians subject to usual exclusions: no Public Trustee visits, no OOP hospitals, no PCHs. Pediatrician visits by those > 14 years of age were excluded. No ER visits included except in ER section.
   
2. Selection of physicians was done from PRDB and limited to Winnipeg physicians with bloc of practice 02 or 11, valid birth date, and practiced in Winnipeg for the whole year.
   
3. Build Solo physician database by identifying TC in medical claims with only 1 (or 2) physician numbers billing. Delete all records with TC of 000. Check for and remove individual physicians (PHYNO) that bill under multiple numbers and remove these. In this study 75 physicians were identified as "solo"' practices.
   
4. The analysis included records with TC= 000 for the calculations but excluded these patients from further analysis. What this meant was the majority of care was determined including 000, so a patient can be rostered to a physician having TEXTCODE=000. Otherwise patients will be falsely rostered to "real" TEXTCODEs. A suggestion for future work would be to use phyno for those records with TC 000. This would force you to make the assumption that physicians not using TCs would be "solo" practices.
   
5. Identify physicians (PHYNO) working in one of the clinics identified. Check if any of these physicians are also identified as "solo" practice physicians.

Based on Menec 2000 report Appendix 1. No code or further notes found.

This report focused on large provider groups. Unfortunately there is no simple way for Manitoba Health to identify physician groups, as there is no group identifier available that indicates whether a physician is affiliated with a group or is working in a solo practice. In order to identify physicians who were practicing out of the same practice location the TC was used. This method may not capture a physician who still files paper claims; the major proportion of physicians in groups file electronically. In this study physician groups had to have at least four general/family practitioners (now referred to as family physicians) and had to have at least four full time equivalent (FTE) physicians. Groups had to have at least 24 four months of data available. Twenty nine groups met these criteria (13 Winnipeg, two Brandon, 14 rural). While 14 of these had some specialists included in the group only ambulatory visits to general/family practitioners [family physicians] were included. At the time of the study only one group consisted mainly of salaried physicians.

Issues

TC DOES NOT ALWAYS identify physicians who practice out of the same practice location and DOES NOT indicate that physician is affiliated with a group practice. Physicians might only be using the same billing service, and using the same billing service may or may not constitute a group practice. This fact also puts limitation on identifying solo physicians. Physicians sharing the same billing service with other physicians can still be solo physicians.

Timing of TC use:


Multiple GPs may have billed simultaneously or on different days under a single TC. If GPs billings overlapped in time or were consecutive billings (one GP is taking over from another; cases of one GP reducing workload, another increasing etc.)

A single GP may bill using multiple TC or without TC at all. Proposed approaches from Profiling Primary care project (Frohlich,2006):


1. Check if the TC dates are all later then no-TC dates then assume TC being there from the beginning of the year. For other TC/no TC combination PHYNOs -- flag them for now.
   
2. If at least 1 TC for each of these PHYNOs is new (StartDate > April 1, 2001), then remove all no-TC claims for these PHYNOs from the consideration.
   
3. Keep Claims submitted through all of TCs for these PHYNOs (they may be excluded later under low volume criteria
   
4. We assumed a group practice CAN NOT have GPs billing with and without TEXTCODE at the same time. Group practice = TEXTCODE with > 1 GP per CODE or group of GPs billing without TEXTCODEs with high % of shared patients (method not finalized).

Manual review of TC names and physician associations should be done for each project using TC to identify solo and group practices.

Separation of walk-in/emergency vs. regular practice may be a problem. MD numbers that can be used by multiple physicians should be resolved or removed.

In the EUSL, Provider Registry data and PRDB there are billing numbers that have missing birth years. These are not real people but represent groups of individuals (see above).

Cautions

The identification of a group practice or clinic is not a trivial matter.

The MDGROUP variable in the medical file has not been updated since early 1990's.

Although TC has been on the MCHP medical claims files since 1994 the associated names and other information about the code in the EUSL flies were only available for specific projects (and for specific sites) until 2004. EUSL files are available online since 2004 with the most recent EUSL files containing historical effective dates for each TC/MD number combination. Since physician numbers can be re-used over time it is not known how this information has been maintained it may be necessary to review historical Provider Registry data sets to separate different physicians with the same MD number.

Solo practice physicians may bill using multiple TCs.

TC may represent a billing service and not a specific site - review the TC names prior to using.

The existence of salaried physicians associated with TC should be determined.