\nOne of the potential shortcomings of the Medicare Part-D prescription database<\/a> is that by definition, it tracks only prescriptions to Medicare beneficiaries over 65 or disabled. The current data for calendar year 2013 is getting\u00a0out of date.\u00a0\u00a0Additionally,\u00a0because the data tracks where the prescriber works and not where the patient lives, small towns that do not have a healthcare professional licensed to prescribe scheduled drugs will not show up in the data at all.\u00a0 Conversely, doctor-heavy cities may be over-represented for\u00a0their populations.\u00a0 While there is little reason to doubt that observations derived from the Medicare population are for the most part generalizable to the population as a whole, I have been searching for a way to demonstrate a useful correlation in a more robust way.\u00a0 Some of the advantages of the Medicare database are that it contains nation-wide data for all prescription drugs and identifies the prescriber by name, provider number, and specialty.\u00a0 I am aware of no other public, or for that matter private database available to match it.\u00a0 If I am mistaken, please let me know and help me obtain access!<\/p>\n
Some systems not set up to track prescribers.<\/b>
\nA major barrier to attempting to make the strongest possible statistical correlation, especially for controlled substances, is that some state opioid tracking systems like Kentucky\u2019s KASPER originally had statutory prohibitions prohibiting the disclosure of the identity of individual providers\u2013 even to licensure or law enforcement organizations\u2013 unless some independent complaint had been lodged.\u00a0 These restrictions, the result of lobbying by provider groups, have in the past prohibited regulatory, law enforcement, or quality-assurance entities such as the the Cabinet for Health and Family Services from looking at dispensing data in a proactive way.\u00a0 KASPER was originally set up to identify patients who were doctor-shopping or otherwise abusing controlled substances, not prescribers. Recent changes in the law allow the Cabinet and law enforcement to proactively look for geographic areas of high prescribing, but I am not aware they can do a whole lot more than that.\u00a0 I find it ironic that anybody can drill down to the individual prescriber level within the Medicare data.<\/p>\n
Preliminary evidence of correlation is fairly good.<\/b> Methods<\/b>. Numerals from 1 to 25 are placed in counties containing cities or towns with the highest rates of city-specific opioid prescriptions per capita. The methodology and limitations of this approach are described in an earlier article<\/a>. Thus the numeral \u20181\u2019 in Lawrence county in southern-most Ohio represents the village of Proctorville which in 2013 had the highest Medicare prescription rate in the state at 6.4 prescriptions per city inhabitant.<\/p>\n Additionally, I placed one or more black dots in each county containing the business addresses of the 25 highest prescribers of opioids in Ohio as also defined elsewhere<\/a>.<\/p>\n Results<\/b>. Highest prescribers.<\/b> Curiously, Columbus, the state\u2019s largest city with double the population of Cincinnati, has a below-average and below-median Medicare prescription rate of 0.19 Rx\/person. Even Cleveland, the second largest city, has a below-average rate of 0.31 Rx\/person. Unravelling the noteworthy differences in the large cities will help define the utility of the Medicare prescription database.<\/p>\n Middletown is in the heart of it after all.<\/b> Conclusion<\/b>. Next look\u2013 Kentucky.<\/b> Peter Hasselbacher, MD
\nSo far, using prescription of opioid analgesics as a model, I am feeling more confident that the Medicare data usefully reflect the situation on the ground.\u00a0 For example, when I looked at the town of Austin, Indiana<\/a>, the locus of a striking epidemic of prescription drug abuse, it was not a surprise to find that several small cities and towns around Austin lead the state in the prescription of Austin\u2019s drug of choice\u2013 Opana.\u00a0 Breakdowns of opioid prescriptions to Medicare patients by city appear to correspond to general public perceptions of the extent of drug abuse and diversion in Kentucky, Indiana<\/a>, and Ohio<\/a>.\u00a0 My recent profile of opioid prescription in Ohio affirms more robustly a correlation of Medicare prescriptions with opioid abuse in the general Ohio population.<\/p>\n
\nIn grappling with its own problem of prescription opioid and heroin use, Ohio constructed its \u201cAutomated RX Reporting System\u201d and makes much information available to the public including a variety of thematic maps<\/a> displaying county-specific information.\u00a0 I borrowed one such map showing the number of opioid doses per capita by county and overlaid it with Ohio data I derived from the Medicare database. [Click map\u00a0see a full sized version.]<\/p>\n![\"ohio-counties-annotated-1000\"](\"http:\/\/www.khpi.org\/blog\/wp-content\/uploads\/2015\/09\/ohio-counties-annotated-1000-135x150.jpg\")
<\/a><\/p>\n
\nTwenty-four of my 25 Medicare \u201chot-spot\u201d cities are co-located in the red or grey-shaded counties identifying those counties with the highest numbers of doses per capita.\u00a0 Only a single one of my hot-spot cities lies in a blue-shaded county with the lowest range of opioid doses per capita\u2013 Noble County.\u00a0 Of note, that small county is situated right in the heart of Ohio\u2019s worst area of prescription drug abuse and is surrounded by hot-spot cities and towns with active\u00a0prescribers of opioids. Both my Medicare data and Ohio\u2019s own data identify southern Ohio and along the I-77 corridor in eastern Ohio as the areas in which opioids are most extensively prescribed. This would not be a surprise to Ohio\u2019s law enforcement and public health authorities.<\/p>\n
\nThe black dots representing the highest individual prescribers lie in counties with the largest cities or in their heavily populated neighboring counties such as those north of Cincinnati. The greater Cincinnati area alone contains 11 of the 25 such highest prescribers!\u00a0 A handful of others are located in southern and eastern, mostly rural Ohio.\u00a0 Of interest is the observation that the urban areas containing the most \u201chighest-prescribers\u201d did not themselves have a hot-spot city as identified by my Medicare criteria. Recognize though, that people in non-urban areas most certainly\u00a0go to the big cities to get many of their prescriptions. Nonetheless,\u00a0cities like Cincinnati, Dayton, Youngstown, and Canton have Medicare prescription rates much higher than the average (0.36) or mean (0.26) rates of prescriptions per person.\u00a0 For example, the average rate in Cincinnati is 0.79 and Dayton 0.67 Rx\/person. With their large populations, the absolute number of prescriptions in the big cities cities adds up fast.<\/p>\n
\nMy interest in Ohio was sparked by the publicized heroin overdose<\/a> of a young woman from Middletown.\u00a0 I was not able to categorize Middletown itself as a hot-spot city, but that city spans Butler and Warren Counties which contain the largest number of highest-prescribers in the state.<\/p>\n
\nThe geographic correlations I describe above are not amenable to formal statistical analysis.\u00a0 However, the overall overlap of these indicators of Medicare prescribing with population-based utilization is in my opinion good enough to proceed with a working assumption that Medicare data provides a useful indicator to the broader use of pharmaceutical products in general provider and patient populations.<\/p>\n
\nClearly, digging more deeply into Kentucky\u2019s experience is what I must do next.\u00a0 Since at least 2011, Kentucky has been publishing Quarterly Trend Reports<\/a> derived from its KASPER system.\u00a0 The reports include both number of doses and per capita use\u2013 by county and 3-digit zip code\u2013 for the most commonly abused opioids. I will ask if annual data for all opioids combined are available. Otherwise I will do a comparison of KASPER and Medicare data drug by drug.\u00a0 When I am even more confident that Medicare data reflects state-wide all-patient prescribing, it will be defensible to include the names of top prescribers in future analyses. Frankly, such names are already in the public domain.\u00a0 It will also be necessary to examine correlations with classes of drugs other than narcotics, but this will require the kinds of extensive drug-utilization databases that only the pharmaceutical companies themselves or large insurance companies seem to have access to!\u00a0 We will have to be creative.\u00a0 Any ideas out there?<\/p>\n
\nPresident, KHPI
\nEmeritus Professor of Medicine, UofL
\nSeptember 20, 2015<\/p>\nShare this:<\/h3>