Google Trends of Radiologists Who Accept Medicare: A Potential Tool for Predicting state Demand

Abstract

Purpose:

To identify and analyze the demand for radiologists who accept Medicare per state from 2004 to 2009, as reflected by volume of Google searches, and to place such demand in context with other available data by state.

Methods:

The number of radiologists who accept Medicare by state was divided by each state’s population to achieve the radiologist density per 10,000 residents. Relative search volume (RSV) for the term “radiologist” was collected from Google Trends from 2004 to 2009. The Radiologist Demand Index (RDI) for each state was then calculated by dividing each state’s RSV by the radiologist density for that state. To standardize values, each state’s RDI was divided by the largest RDI to generate the Relative Radiologist Demand Index (RRDI). Utilization of medical imaging per 1000 Medicare beneficiaries in each state, overall health of a population in each state, and percentage of the population enrolled in Medicare in each state were used to compare trends with the RRDI.

Results:

West Virginia had the greatest curiosity about radiologists who accept Medicare (as represented by proportion of Google searches) (RSV=100), followed by Mississippi (RSV=95), and Arkansas (RSV=87). Oregon demonstrated the lowest level of curiosity about radiologists who accept Medicare, by having the lowest proportion of google searches (RSV=43), followed by Vermont (RSV=49), California (RSV=50), and Colorado (RSV=50). The highest radiologist densities per population were found in Montana, D.G, and Wyoming (3.25,1.56,1.11, respectively). The lowest radiologist densities were found in Oklahoma, Texas, and Utah (0.4, 0.4, 0.41,0.41, respectively). The RRDI was greatest in Louisiana (100), Arkansas (94.8), and Texas (86.3), and smallest in Montana (10.6), D.G (17.7) and Wyoming (28.4). Positive trends between utilization of medical imaging per 1000 Medicare beneficiaries and state overall health and the RRDI were recognized. No trend between each state’s RRDI and percentage of population enrolled in Medicare was noted.

Conclusion:

Imaging studies performed, an indirect measure of demand, showed trends with RRDI. Higher RRDI and imaging per 1000 Medicare beneficiaries trended with lower health scores for a state’s general population. RRDI may be a useful tool reflecting each state’s demand for radiologist who accepts Medicare.

Introduction

Google Trends was launched in 2006 as a way to quantify and analyze the popularity of search inquiries both geographically and temporally. Shortly after its launch, researchers discovered the data could be used for a variety of outlets such as predicting the stock market, disease outbreaks, and election results.^1–5 Recently, Google Trends has been used to predict the demand for plastic surgeons in the United States and Austria with the goal of informing surgeons of their economic prospects based on geography.^6,7

Diagnostic imaging accounts for 10% of total US healthcare expenses.⁸ From 2000 to 2007, utilization of imaging increased 70% for Medicare beneficiaries compared to all other physician services at 40%.⁹ Utilization of radiological services peaked around 2008, then sharply declined largely due to the Deficit Reductions Act (DRA) in 2005.¹⁰ The DRA was designed to cut increasing imaging costs for Medicare and therefore had an enormous impact on utilization. For example, a study investigating Medicare radiology claims per thousand enrollees showed a steady increase in claims from 2003 until the peak of 4422 claims per thousand enrollees per year in 2008.¹¹ Subsequently, there was a sharp decrease that remained steady from 2009 to 2016 at around 3410 claims per thousand enrollees per year. Ultimately, the use of medical imaging per Medicare enrollees is liable to fluctuate due to government policy.

Medicare is a program that began in 1966 under President Lyndon B. Johnson as a way to provide nationalized health care to all Americans 65 and older. Currently, eligibility requirements extend to those younger than 65 who are disabled and all people who are on dialysis. Medicare is funded through a tax placed on both employees and employers. In 2019, Medicare spending increased 6.7% to 799.4 billion dollars covering a total of 64 million Americans¹² Of this population about 86% were eligible by age and 14% were covered due to disability or chronic renal disease. By 2030, all of the baby boomer generation will be over 65 meaning around 70 million people will be covered due to age.¹³ The number of physicians who accept Medicare will have to grow to meet the demand of the United States’ aging population. To our knowledge there is no data available in the literature describing the demand of U.S. board-certified radiologists who accept Medicare using Google Trends.

The purpose of this study was to investigate trends between the Relative Radiologist Demand Index (RRDI), and various factors including population on Medicare, utilization of medical imaging, and overall health of the population as measured by United Health Foundation report, to assess its usefulness as a tool to predict demand for radiologists who accept Medicare.

Methods

The RSV (relative search volume) for the term “Radiologist” was collected from Google Trends from January 1,2004 to December 31, 2019. The term “Radiologist Medicare” was also used but did not add any value to the existing data. The terms were chosen as a way to assess each state’s interest in searching for radiologists who accept Medicare. This 15-year time frame extends from the beginning of Google Trends data to the latest full year at the time of the data collection. This data was reported by state and standardized from 0 to 100 based upon proportion of total searches.^14,15 Each state’s population, including the District of Colombia, was collected using the US Census projected population as of July 1,2019. The number of radiologists per state that accept Medicare was tallied using the 2020 Medicare publicly accessible database called the “Physician Compare National Downloadable file”.

The Radiologist Demand Index (RDI) was used to determine public interest as measured by Google searches relative to the concentration of radiologists who accept Medicare. To standardize the data, each state’s RDI was individually divided by the largest RDI (Louisiana at 206.6) and multiplied by 100 to create the RRDI. A choropleth map was created to illustrate the RRDI for each state.

$$\mathit{Radiologist}\mathit{Demand}{\mathit{Index}}_{\mathit{State}}=\frac{\mathit{Relative}\mathit{Search}{\mathit{Volume}}_{\mathit{State}}}{\mathit{Radiologist}{\mathit{Density}}_{\mathit{State}}}$$

The United Health Foundation rates each state on a variety of categories including environment, policy, and medical care and creates a ranking of states by overall health. The utilization of medical imaging per Medicare beneficiaries, the health ranking of each state’s population, and the percentage of population on Medicare per state were acquired.^16–18 Each of these were compared to the RRDI for each state to assess for trends with the RRDI.

A line graph ranking the states from highest to lowest on health, according to the United Health Foundation 2019 Annual Report, was constructed. Imaging procedures per 1000 Medicare beneficiaries, RRDI, and percentag

Results

West Virginia had the largest RSV for the term “radiologist” with a value of 100, followed by Mississippi (95), Arkansas (87), Tennessee (85), and South Dakota (85) (Table 1). Oregon had the smallest RSV at 43, followed by Vermont (49), California (50), Colorado (50).

TABLE 1.

State ranking of demand for radiologists who accept Medicare including relative search volume, radiologist density, and Medicare Part B all imaging procedures per 1000 beneficiaries.

Rank, Radiologist Demand Index	State Name	Radiologists that accept Medicare	State Population	Radiologist Density (Radiologists per 10,000 people)	Google Relative Search Volume	Radiologist Demand Index	Relative Radiologist Demand Index	Medicare Part B All imaging procedures per 1000 beneficiaries
1	Louisiana	189	4,648,794	0.41	84	206.6	100	3882.48
2	Arkansas	134	3,017,804	0.44	87	195.9	95	3232.59
3	Texas	1171	28,995,881	0.40	72	178.3	86	3068.01
4	Oklahoma	160	3,956,971	0.40	71	175.6	85	3428.51
5	West Virginia	103	1,792,147	0.57	100	174.0	84	2955.56
6	North Dakota	33	762,062	0.43	75	173.2	84	2319.1
7	Mississippi	173	2,976,149	0.58	95	163.4	79	3618.47
8	Alabama	224	4,903,185	0.46	74	162.0	78	2232.12
9	Nevada	141	3,080,156	0.46	74	161.7	78	2930.92
10	Utah	131	3,205,958	0.41	65	159.1	77	2348.64
11	Tennessee	367	6,829,174	0.54	85	158.2	77	2392.23
12	North Carolina	479	10,488,084	0.46	72	157.6	76	1955.13
13	Georgia	462	10,617,423	0.44	68	156.3	76	2481.87
14	Florida	1039	21,477,737	0.48	75	155.0	75	3604.44
15	Arizona	352	7,278,717	0.48	72	148.9	72	2890.01
16	South Carolina	264	5,148,714	0.51	76	148.2	72	2992.06
17	Indiana	376	6,732,219	0.56	72	128.9	62	3312.52
18	Kentucky	292	4,467,673	0.65	81	123.9	60	3253.62
19	Missouri	344	6,137,428	0.56	69	123.1	60	3013.71
20	New Mexico	110	2,096,829	0.52	64	122.0	59	2785.66
21	Illinois	665	12,671,821	0.52	64	122.0	59	1673.25
22	California	1644	39,512,223	0.42	50	120.2	58	2934.84
23	South Dakota	64	884,659	0.72	85	117.5	57	2573.3
24	Maine	73	1,344,212	0.54	63	116.0	56	2620.93
25	Michigan	534	9,986,857	0.53	62	116.0	56	3888.95
26	Idaho	96	1,787,065	0.54	62	115.4	56	1085.1
27	Nebraska	126	1,934,408	0.65	74	113.6	55	3068.36
28	Ohio	670	11,689,100	0.57	64	111.7	54	2419.8
29	Colorado	260	5,758,736	0.45	50	110.7	54	2423.44
30	New York	1241	19,453,561	0.64	67	105.0	51	3539.52
31	Kansas	187	2,913,314	0.64	65	101.3	49	2020.26
32	Lowa	204	3,155,070	0.65	65	100.5	49	3020.95
33	Washington	412	7,614,893	0.54	53	98.0	47	2393.66
34	Connecticut	245	3,565,287	0.69	67	97.5	47	2859.61
35	Minnesota	347	5,639,632	0.62	59	95.9	46	2230.49
36	Wisconsin	422	5,822,434	0.72	69	95.2	46	2085.6
37	Virginia	492	8,535,519	0.58	53	91.9	45	2926.4
38	Pennsylvania	836	12,801,989	0.65	59	90.3	44	2868.31
39	New Jersey	631	8,882,190	0.71	64	90.1	44	2155.09
40	Rhode Island	65	1,059,361	0.61	55	89.6	43	2650.79
41	Maryland	414	6,045,680	0.68	61	89.1	43	2642.25
42	Alaska	60	731,545	0.82	65	79.3	38	2861.21
43	New Hampshire	99	1,359,711	0.73	53	72.8	35	2580.79
44	Oregon	251	4,217,737	0.60	43	72.3	35	2468.38
45	Delaware	83	973,764	0.85	60	70.4	34	2966.78
46	Massachusetts	571	6,892,503	0.83	53	64.0	31	2748.33
47	Vermont	48	623,989	0.77	49	63.7	31	2020.9
48	Hawaii	143	1,415,872	1.01	63	62.4	30	1523.18
49	Wyoming	64	578,759	1.11	65	58.8	28	2420.23
50	District of Columbia	110	705,749	1.56	57	36.6	18	1558
51	Montana	347	1,068,778	3.25	71	21.9	11	2648.47

California had the highest number of radiologists who accepted Medicare with 1644. North Dakota had the smallest number of radiologists who accepted Medicare at 33. When correcting for state population, Montana had the highest density of radiologists who accept Medicare with 3.25 per 10,000. After Montana, the states in which the density of radiologists who accept Medicare were next highest in descending order were D.C (1.56), Wyoming (1.11), Hawaii (1.01), and Delaware (0.85). Oklahoma and Texas had the lowest density of radiologists who accept Medicare per population with 0.4 per 10,000. This was followed by Louisiana (0.41), Utah (0.41), and California (0.42).

The state with the highest public interest, measured by Google searches, relative to the concentration of radiologists who accept Medicare (RDI) is Louisiana. The standardization calculation results in a Louisiana RRDI value of 100 (Fig 1). In descending order of RRDI (an analogue for demand for radiologists who accept Medicare), Louisiana was followed by Arkansas (94.8), Texas (86.3), Oklahoma (85.0) and West Virginia (84.2). Montana has the smallest RRDI with a value of 10.6. This was followed in ascending order of RRDI by DC (17.7), Wyoming (28.4), Hawaii (30.2), and Vermont (30.8).

FIG 1.

The relative radiologist demand index (RRDI) by state. The darker the shade of color the higher the RRDI value (higher demand). The lighter the shade of color the lower the RRDI value (lower demand).

We observed a trend that states with higher RRDIs (demand) had higher utilization of imaging per state by Medicare beneficiaries. States with lower RRDIs had lower utilization of imaging per state by Medicare beneficiaries (Fig 2).

FIG 2.

Medicare Part B All Imaging procedures per 1000 beneficiaries, Relative Radiologist Demand Index (RRDI) per state and Total enrollment medicare percent of resident population with respect to overall health grade of population as measured by the United Health Foundation’s 2019 annual report The x-axis is the ranking of states by their overall health grade from most to least healthy as according to the United Health Foundation. The orange line represents the Medicare Part B All Imaging Procedures per 1000 beneficiaries and is scaled using the left y-axis ranging from 0 to 4000. The blue line represents the Relative Radiologist Demand Index and is scaled using the right y-axis ranging from 0 to 100. The black line represents the Total Enrollment Medicare Percent of Resident Population and is scaled using the right y-axis ranging from 0 to 100.

States with higher RRDIs trended with lower overall health scores, according to the United Health Foundation 2019 Annual Report, compared to states with lower RRDIs. (Fig 2). States with higher utilization of medical imaging trended with lower scores of overall health on the United Health Foundation 2019 Annual Report.

No trend was observed between the percentage of the population that is on Medicare with the RRDI (Fig 2). The percentage of the population on Medicare ranged from 12 to 24.1 and showed random variation when compared to the RRDI.

Discussion

Our results show that the majority of states demonstrated similar trends between the imaging procedures ordered per 1000 Medicare beneficiaries and overall health of the population using the RRDI as a measure of interest. No such similar trend was observed for the percentage of the population on Medicare and the RRDI.

We observed similar trends in imaging procedures ordered per 1000 Medicare beneficiaries and the RRDI. Utilization of imaging procedures has been used in a model to predict demand for radiologists’ services.¹⁹ The positive trend our study observed may suggest there is a potential to use the RRDI as a measure to estimate imaging study demand in the Medicare population (Table 1, Fig 2). For example, states with high RRDIs such as Louisiana and Arkansas, the number of imaging procedures per 1000 beneficiaries in 2018 was 3882 and 3232 respectively (Table 1, Fig 2). In low RRDI states such as Montana and D.C., the number of imaging procedures per 1000 beneficiaries in 2018 was 2648 and 1558 respectively (Table 1, Fig 2). These results are further supported by previous literature that showed positive correlation between search volume of various plastic surgery procedures using Google Trends and actual procedures performed.²⁰

However, the data from some states did not follow this trend of imaging studies performed and RRDI values, such as Montana. This may indicate potential confounders such as differences in imaging referral patterns or effects related to differences in health literacy, socioeconomic status, adherence with health care guidelines, or other unknowns.

There were similar trends for RRDI values and overall health in the population as measured by United Health Foundation (DC is not included in their ranking list) (Fig 2). The five states with the highest demand for radiologists who accept Medicare (LA, AK, TX, OK, WV) all fall in the bottom half of the country for overall health, based on the United Health Foundation 2019 Annual Report. With the exception of Texas, these states rank in the bottom 10^th percentile for health. Interestingly, the states with the lowest demand for radiologists who accept Medicare (MO, WY, HI, VT, MA) are among the top 25 healthiest states in the United States with Vermont being rated as the healthiest.

The percentage of the population that is on Medicare did not show any trend with the RRDI. For example, Louisiana with the highest RRDI value has 17.7% population on Medicare, and Montana with lowest RRDI values has 20.5% of population on Medicare (Fig 2). This suggests that a higher population on Medicare does not necessarily reflect an increased need for imaging studies.

This study had several limitations. First, patients may find their radiologists through other sources such as referring providers, word of mouth, or by using other internet search platforms, which would not be captured in our study. Additionally, other confounding factors that might impact health care management including but not limited to socioeconomic status, location, health literacy, and government policy are not studied. The present study does not encompass the nuances of individual access to medical care.

Conclusion

RRDI is a tool derived from Google Trends that may be useful in the estimation of the number of imaging studies in the Medicare population, therefore serving as an indirect measure of demand for Radiologists who accept Medicare. Interestingly, we observed a higher value of RRDI and number of imaging studies per 1000 Medicare beneficiaries with overall poorer health of a state’s general population, however, no trend was noted for total Medicare population.

This study provides government officals and radiologists another potential tool for gaining insights aboiut where the greatest needs exist for radiologists who accept Medicare in the United States.