34:["$","title",null,{"children":"Trends in Treatment | Caroline Cullinan | Observable"}] 3c:T671,html` Substance use disorders present a plethora of health-related concerns for not only those who use substances, but also for society at large. According to the Lancet Global Burden of Disease (GBD) Resource Center, 11.8 million people are killed each year as a result of tobacco, alcohol, and illicit drug substance use disorders. The loss of such lives has tragic ramifications for both families and greater communities. As such, it should come to no surprise that, globally, behavioral health departments are looking to evaluate their Substance Use Detox Programs for future improvement.

This Observable notebook outlines data visualizations for 1) exploratory analysis and 2) K-Prototype Clustering analysis of provided Substance Use Detox Program data for the 2015-2015 fiscal year of an unnamed medium-sized country.

While the data-providing client who hired this analysis was primarily concerned with understanding if certain groups of people were staying in the Substance Use Detox Program longer than the expected 5 days, this analysis also considers the success of people enrolled in such programming. In doing so, a new main research question has been propose:

Who is staying in treatment for more or less than the expected number of days, and how successful are these people in reducing their frequency of substance use?

In answering this question, the client may be better equipped to improve their services and ensure better health outcomes for people enrolled in the Substance Use Detox Program.

`3d:T7b8,html` Section 1.0 of this Observable notebook outlines visualizations pertaining to the exploratory analysis of this research. More specifically, a comparison of days spent enrolled in treatment and substance use reduction success is made at both the aggregated and disaggregated levels (i.e., with regard to different demographic characteristics). The patterns that were qualitatively observed in this exploratory analysis ultimately called for the K-Prototype Clustering analysis outlined in Section 2.0. Section 2.0 of this Observable notebook outlines visualizations pertaining to the K-Prototype Clustering analysis that was conducted as a result of the qualitatively observed patterns in the exploratory analysis. Specifically, the visualizations in this section illustrate the discovered ‘clusters’ of different people in treatment, and the different demographic attributes associated with each cluster. In addition, visualization in this section also illustrates the median number of days spent enrolled in treatment in relation to the median relative change in substance use frequency per cluster. In Section 3.0 an interpretation of the overall analysis is presented, and finally, in Section 4.0, recommendations for future data collection and analysis are made

Please see the Appendix for access to the code behind this analysis and for a complete explanation of assumptions made, data metrics, and analysis methods.

Many alternative visualization options have been made available for each figure. For each figure, please be sure to interact with the buttons located above each chart. In doing so, you can modify visualization elements, and for some of the figures you can also switch between different metrics.

`3e:T453,{ const plot = Plot.plot({ x: { padding:0.5, label:"Days Enrolled" }, y: { padding:0.5, label:"Number of People Enrolled" }, marks:[ Plot.gridY({opacity: 0.1}), Plot.barY(data_raw, Plot.groupX({y: "count"}, {x: "enrollment_v_target", fill:'#44BB99', width:50, sort: {y: "x", reverse: true, limit: 10}})), Plot.ruleY([0]) ], }) const plot2 = Plot.plot({ x: { padding:0.5, }, y: { padding:0.5, label:"Number of People Enrolled" }, marks: [ Plot.gridY({opacity: 0.1}), Plot.barY(data_raw, Plot.groupX({y: "count"}, {x: "program_success", fill:'#44BB99', sort: {y: "x", reverse: true, limit: 10}})), Plot.ruleY([0]) ], // caption: "where 'success' indicates that there has been a reduction in substance use (i.e., substance use has neither increased nor stayed the same)" }) return html`

${[plot, plot2]}

` }3f:T46a,Plot.plot({ y: { label: "Proportion (%)", tickFormat: d=> d*100 }, x: { label: "", padding:0.5, }, facet: { data: showFacet?data_raw:null, fx: xVar==="enrollment_v_target"?"program_success":"enrollment_v_target", label: "" }, color: { legend:true, //range: ["#E69F00", "#56B4E9", "#009E73", "#F0E442", "#D55E00" , "#CC79A7", "#0072B2"] range: tol.QualLight }, marks: [ // Plot.stackY( Plot.gridY({opacity: 0.1}), Plot.barY( data_raw, Plot.stackY(Plot.groupX({y: "proportion"}, { x: xVar, fill: showVar, // offset: "normalize", offset: normalize?"normalize":null, } ) ) ), // ), , Plot.ruleY([0]) ], title:"Title of my chart", caption: showFacet===false&&xVar=='enrollment_v_target'? '':"where 'success' indicates that there has been a reduction in substance use (i.e., substance use has neither increased nor stayed the same)" })40:T447, { const showVariables = ['male_proportion', 'female_proportion', 'BIPOC_proportion', 'white_proportion', 'disabled_proportion', 'not_disabled_proportion', 'aged_20-30_proportion', 'aged_31-50_proportion'] const showData = k_explore_tidy.filter(d=>showVariables.includes(d.variable)) const plot = Plot.plot({ width: 1200, grid:true, y: { label: "Proportion (%)", // tickFormat: d=> d*100 }, // x: { // label: "" // }, facet: { data: showData, // y: (d) => Math.floor(d["variable"] / 10), // x: (d) => d["variable"] % 10, x: (d) => d["variable"], domain: ['male_proportion', 'female_proportion', 'BIPOC_proportion', 'white_proportion', 'disabled_proportion', 'not_disabled_proportion', 'aged_20-30_proportion', 'aged_31-50_proportion'] // label: "" }, color: { legend:true, type: "categorical", range: tol.QualLight }, marks: [ // Plot.stackY( Plot.barY(showData, {x: "cluster", y:"value", fill: "cluster", sort: {y: "x", reverse: true, limit: 10}}) // ), , Plot.ruleY([0]) ] }) return plot } 41:T45e, { const showVariables = ['aged_51-70_proportion', 'alcohol_proportion','heroin_proportion', 'methamphetamine_proportion','marijuana_proportion', 'cocaine_crack_proportion', 'oxycodone_oxycontin_proportion'] const showData = k_explore_tidy.filter(d=>showVariables.includes(d.variable)) const plot = Plot.plot({ width: 1200, //inset: 5, grid:true, y: { label: "Proportion (%)", // tickFormat: d=> d*100 }, // x: { // label: "" // }, facet: { data: showData, // y: (d) => Math.floor(d["variable"] / 10), // x: (d) => d["variable"] % 10, x: (d) => d["variable"], domain: ['aged_51-70_proportion', 'alcohol_proportion','heroin_proportion', 'methamphetamine_proportion','marijuana_proportion', 'cocaine_crack_proportion', 'oxycodone_oxycontin_proportion'] // label: "" }, color: { //legend:true, type: "categorical", range: tol.QualLight }, marks: [ // Plot.stackY( Plot.barY(showData, {x: "cluster", y:"value", fill: "cluster", sort: {y: "x", reverse: true, limit: 10}}) // ), , Plot.ruleY([0]) ] }) return plot } 42:T7a9, { const showVariables1 = ['treatment_days'] const showData1 = k_explore_tidy.filter(d=>showVariables1.includes(d.variable)) const plot = Plot.plot({ // grid:true, y: { label: "# of Days", // tickFormat: d=> d*100 }, x: { label: "" }, // facet: { // data: showData, // // y: (d) => Math.floor(d["variable"] / 10), // // x: (d) => d["variable"] % 10, // x: (d) => d["variable"], // domain: ['male_proportion', 'female_proportion', 'BIPOC_proportion', 'white_proportion', 'treatment_days'] // // label: "" // }, color: { legend:true, type: "categorical", range: tol.QualLight, label: "cluster" }, marks: [ Plot.gridY({opacity: 0.1}), // Plot.stackY( Plot.barY(showData1, {x: "cluster", y:"value", fill: "cluster", sort: {y: "x", reverse: true, limit: 10}}) // ), , Plot.ruleY([0]) ] }) const showVariables2 = ['change_freq_relative'] const showData2 = k_explore_tidy.filter(d=>showVariables2.includes(d.variable)) const plot2 = Plot.plot({ // grid:true, y: { label: "Change in Relative Substance Use (%)", // tickFormat: d=> d*100 type: "symlog", ticks: [-100, -75,-50, -25, 0, 25, 50, 75, 100, 1000], grid: true }, x: { label: "" }, // facet: { // data: showData, // // y: (d) => Math.floor(d["variable"] / 10), // // x: (d) => d["variable"] % 10, // x: (d) => d["variable"], // domain: ['male_proportion', 'female_proportion', 'BIPOC_proportion', 'white_proportion', 'treatment_days'] // // label: "" // }, color: { // legend:true, type: "categorical", range: tol.QualLight, }, marks: [ Plot.gridX({opacity: 0}), // Plot.stackY( Plot.barY(showData2, {x: "cluster", y:"value", fill: "cluster", sort: {y: "x", reverse: true, limit: 10}}) // ), , Plot.ruleY([0]) ] }) return html`

${[plot, plot2]}

` } 43:T12d9,html`Cluster 0:
This cluster is composed of predominantly white (76% white) males (70% male) without disabilities (8% disabilities). People in this group are predominantly middle-aged (i.e., 75% between the ages of 31 and 50 years old). People in this cluster also predominantly use alcohol (64% alcohol use), although they also use methamphetamine (20% methamphetamine use) and heroin (9% heroin use) to lesser extents. These people are usually in treatment for about 1-20 days (with a median number of treatment days = 5) , which is longer than Cluster 1 people (anywhere from about 1-15 days with a median number of treatment days = 3) and less than Cluster 3 people (anywhere from about 20-40 days with a median number of treatment days = 29). These people also tend to be relatively successful at decreasing their substance use after treatment (i.e., they benefit the second most from treatment compared to other clusters, with a median relative substance use frequency reduction of 37%). Notably, however, when compared to Cluster 1, people in Cluster 0 make up less of the people who are in treatment for less than or equal to 5 days (i.e., the amount of time expected in treatment). While some people in this cluster spend more than 5 days in treatment, they represent less people who spend more than 5 days in treatment when compared to Cluster 3.

Cluster 1:
This cluster is composed of predominantly white (78% white) males (70% male) without disabilities (9% disabilities). People in this group are predominantly NOT middle-aged, but instead younger (i.e., 25% between the ages of 20 and 30 years old) and older (i.e., 67% between the ages of 51 and 70 years old). People in this cluster predominantly use alcohol (69% alcohol use), although they also use methamphetamine (15% methamphetamine use) and heroin (10% heroin use) to lesser extents. These people are usually in treatment for about 1-15 days (with a median number of treatment days = 3) , which is shorter than both Cluster 0 people (anywhere from about 1-20 days with a median number of treatment days = 3) and Cluster 3 people (anywhere from about 20-40 days with a median number of treatment days = 29). People in Cluster 1 make up the most of the people who complete treatment in 5 or less days (84% of Cluster 1 completes treatment in <= 5 days). Regardless, people in this cluster are not the most successful in terms of reducing their median relative frequency of substance use (median relative substance use frequency reduction of 25%).

Cluster 2:
This cluster represents outliers. Notably, however, this cluster has the most number of people who are disabled (33% disabilities). In addition, people in this cluster also have a median relative substance use frequency increase of 1400%, meaning that they ended up using substances even more after treatment. Interestingly, in the exploratory analysis phase of this project, it was determined that some people changed substances during treatment (e.g., they went from primarily using methamphetamine to primarily using alcohol), some people increased their frequency of substance use after treatment (i.e., they started to use substances even more after treatment), and some people even changed substances during treatment AND increased their frequency after treatment. This cluster represents these people, and it is important to approach these cases with a more nuanced lens (Using qualitative methods, a case-by-case basis could be used to approach these outlier patients, and a series of interviewing could take place in order to better understand how their experiences during treatment could have contributed to such changes in substance use and/or increase in substance use frequency.)

Cluster 3:
This cluster is composed of predominantly BIPOC (68% BIPOC) males (70% male) that are younger (i.e., 33% between the ages of 20 and 30 years old) and middle-aged (i.e., 67% between the ages of 31 and 50 years old). Compared to Cluster 0 and Cluster 1, a higher proportion of people in Cluster 3 have disabilities (25% disabilities). People in this cluster predominantly use alcohol (47% alcohol use), although they also use methamphetamine (32% methamphetamine use) and heroin (16% heroin use) to lesser extents. With a median number of treatment days equal to 29, people in this cluster are also usually in treatment for much longer than the other clusters. With that said, 0% of people in this cluster are in treatment for 5 days or less. Regardless of this fact, the median relative change in substance use frequency for Cluster 3 people is -67%, making Cluster 3 patients the most successful patients who undergo treatment (i.e., these patients experience the biggest reductions in relative substance frequency use). `44:T472,html`Despite being made up of higher proportions of historically marginalized groups (i.e., BIPOC people and people with disabilities) who may face even further institutional barriers to successfully receiving substance use treatment, Cluster 3 is the cluster where people are most successful in reducing their relative frequency of substance use. Simultaneously, Cluster 3 is also the cluster which represents people who spend the most amount of time enrolled in program treatment. Contrastingly, and not considering the outlier Cluster 2, Cluster 1 represents the cluster where people spend the least amount of time in treatment, while also having the smallest reductions in relative substance use frequency. Interestingly, and when compared to Cluster 1, Cluster 0 represents a cluster where people spend a bit more time in treatment, and slightly better reductions in relative substance use frequency result.

Thus, more time in treatment may result in greater reductions in relative frequency of substance use, and less time in treatment may actually correlate with less success in reducing substance use in the long run.`45:T49b,html`In answering the original research question:

Who is staying in treatment for longer than the expected 5 days?

100% of people in Cluster 3 stay in program treatment for longer than the expected 5 days. 43% of Cluster 0 people stay in treatment for longer than the expected 5 days, and only 16% of Cluster 1 people stay in treatment for longer than the expected 5 days.

In answering the new research question:

Who is staying in treatment for more or less than the expected number of days, and how successful are these people in reducing their frequency of substance use?

100% of people in Cluster 3 stay in program treatment for longer than the expected 5 days, and these people reduce their relative frequency of substance use by a median value of 67%. 43% of Cluster 0 people stay in treatment for longer than the expected 5 days, and they reduce their relative frequency of substance use by a median value of 37%. Finally, only 16% of Cluster 1 people stay in treatment for longer than the expected 5 days, and they only reduce their relative frequency of substance use by a median value of 25%

`46:T65b,html`Given the outcomes of this overall analysis, which indicate that more time in treatment may result in greater reductions in relative frequency of substance use and less time in treatment may actually correlate with less success in reducing substance use in the long run, the following recommendations for immediate data collection and analysis are as followings:

1. Qualitatively investigate Cluster 2 program enrollment. On a case-by-case basis, investigate these outlier people with a series of interviews in order to better understand how their experiences during treatment could have contributed to such changes in substance use and/or increase in substance use frequency.

2. Further explore the relationship between program enrollment time, the number of days spent in treatment, and the reduction in substance use relative frequency.

3. Collect more data relevant to people’s access to transportation, familiar support, employment, etc. in order to enrich any further immediate analysis.

4. Conduct more clustering analysis on the expanded data

5. Ultimately, however, it is also recommended that the Behavioral Health Department of this unnamed middle-sized country develops a metric for success in terms of substance frequency reduction post treatment enrollment. With a clearer metric of success, the most appropriate number of days expected to be in treatment can be determined. Given this preliminary analysis, it is evident that people may not be able to benefit from treatment the most if they are only enrolled in the programming for 5 days or less.

`47:Tdb3,html`1.0 Assumptions

In conducting this analysis, the following assumptions have been made:

* The client is ultimately interested in conducting this analysis to improve their services and the ultimate health of people who seek their services.

* People who enter program treatment and exit program treatment on the same day complete 1 full day of programming.

* Given that the frequency of substance use is recorded over a 30 day period, it is assumed that substance use occurs only once daily. For example, if somebody entered program treatment with a substance use frequency of 30, it would be assumed that this person used substances once a day for 30 days, i.e., their relative frequency of substance use would be 100%.

2.0 Data

All data used to conduct this analysis was provided by the client, the Behavioral Health Department of an unnamed middle-sized country.

2.1 Data Metrics

Important to this analysis are the following data metrics:

Number of Days in Treatment:

calculated by counting the number of days between program entry date and program exit date, but where program entry and exit on the same day count as 1 full day of programming

Change in Relative Substance Frequency Use:

((Exit Relative Substance Use Frequency - Start Relative Substance Use Frequency) / Start Relative Substance Use Frequency) ) * 100

where: Relative Substance Use Frequency = Substance Use Frequency / 30 (days)

3.0 Analysis

In order to analyze this data, the Python and JavaScript programming languages were used. Observable Plot was used to compile all analysis and visualizations in an Observable notebook.

3.1 Exploratory Analysis

With Python, data was wrangled, analyzed, summarized, and visualized. Additional exploratory data analysis was also performed in order to visually identify relationships between different demographic characteristics, the number of days enrolled in treatment, and program success (i.e., whether or not a person reduced the frequency of their substance use after treatment). JavaScript was used to create the accompanying data visualizations that outline this analysis.

3.2 K-Prototype Clustering

With Python, a K-Prototype clustering analysis was performed in order to determine clusters of people enrolled in treatment with similar characteristics and predict who 1) would likely be enrolled in treatment for more or less than 5 days, and 2) who would most likely reduce their frequency of substance use the most after treatment. The elbow plot method was used to determine the optimal number of K clusters. To conduct this analysis, K = 4 clusters (resulting in a CV F1 score of 0.83 (i.e., good)) and the ‘Huang’ initialization method were used. Following analysis, cluster outcome data was further wrangled, analyzed, summarized, and visualized in Python. JavaScript was used to create the accompanying data visualizations that outline this analysis.

For more information on the K-Prototype Clustering analysis method used, click here

Click here to access the code for the analysis.

`48:[{"id":"9fb8741cd62f475eb3b89d43980589c07f3750d976fd64d556e7c6003759747159f80ebe3c00b2faf8e6e1d5b5268b5865cdf3fca6aaf0c413f2380824ab851e","href":"https://static.observableusercontent.com/files/9fb8741cd62f475eb3b89d43980589c07f3750d976fd64d556e7c6003759747159f80ebe3c00b2faf8e6e1d5b5268b5865cdf3fca6aaf0c413f2380824ab851e","name":"k_explore.csv","type":"text/csv","lastModified":1680554354231.923,"size":3351},{"id":"90bc81292a4fbab6bc18deda2d3b5de879ce8a686fbc8dc72b651e3efc910a5dc39d8cff860dbe34b35848fd7877f59510c9f96318d16b8b5e76bee1e290a304","href":"https://static.observableusercontent.com/files/90bc81292a4fbab6bc18deda2d3b5de879ce8a686fbc8dc72b651e3efc910a5dc39d8cff860dbe34b35848fd7877f59510c9f96318d16b8b5e76bee1e290a304","name":"data_raw.csv","type":"text/csv","lastModified":1680554355945.431,"size":168465},{"id":"bdadfb40bcdf948ce62e57a37d2738cc12d780f7f65f1a18c1a104e566e7fef6abf829a47dee571a18e7173247b484b5b6f4eb36cc57db2a9d44b871a6b723e1","href":"https://static.observableusercontent.com/files/bdadfb40bcdf948ce62e57a37d2738cc12d780f7f65f1a18c1a104e566e7fef6abf829a47dee571a18e7173247b484b5b6f4eb36cc57db2a9d44b871a6b723e1","name":"k_explore_tidy.csv","type":"text/csv","lastModified":1680557983232.327,"size":7405}] 38:{"id":"d4ed36a9e390d04d","slug":"substance-use-detox-program-trends","title":"Trends in Treatment","thumbnail":"d39704c831ff8e234a8f6a5ef646db6b15ba663cb24a2bb5a2696c0b4efa9aa1","access_level":"public","create_time":"$D2023-04-02T16:31:27.167Z","update_time":"$D2023-07-07T17:24:33.723Z","view_time":null,"first_public_time":"$D2023-04-02T16:31:27.167Z","has_unpublished_edits":null,"fork":false,"stars":0,"starred":false,"owner":{"type":"team","id":"b783a0f584d21589","login":"carolinecullinan","name":"Caroline Cullinan","avatar_url":"https://avatars-next.observableusercontent.com/8b89c23b31cfac4c9e38782347b8515d60849281a99fcc3ec016646c0f466bae","tier":"starter_2024","bio":"Participatory Systems | Community-Driven Development","followed_by_you":false,"follows_you":false},"version":595,"publish_version":null,"role":null,"authored":false,"authors":[{"type":"individual","id":"3252c0480a163254","login":"carolinecullinan","name":"Caroline Cullinan","avatar_url":"https://avatars-next.observableusercontent.com/36537301c4253b6e5b85cebe4262781beb533d83fa0dfc319d1d1463d6166cf9","tier":null,"bio":"","followed_by_you":false,"follows_you":false}],"members":[{"type":"individual","id":"3252c0480a163254","login":"carolinecullinan","name":"Caroline Cullinan","avatar_url":"https://avatars-next.observableusercontent.com/36537301c4253b6e5b85cebe4262781beb533d83fa0dfc319d1d1463d6166cf9","tier":null,"role":"editor"}],"comments":[],"custom_thumbnail":false,"trashed":false,"lapsed":null,"subscribed":null,"forks":0,"collection":null,"type":"notebook","body":{"titling":"auto","title":"Trends in Treatment","cells":[{"id":12,"value":"html`

Trends in Treatment

Identifying Trends in Substance Use Detox Programming Patient Data for the 2015-2016 Fiscal Year

Prepared for Resource Development Associates (RDA) by Caroline Cullinan
`","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":20,"value":"$3c","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":401,"value":"$3d","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":458,"value":"html`

Days Enrolled in Treatment and Substance Use Reduction Success

A comparison of time spent enrolled in program treatment and substance use reduction success for all clients in the 2015-2016 fiscal year`","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":521,"value":"$3e","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":526,"value":"htl.html`

Where 'success' indicates that there has been a reduction in substance use (i.e., substance use has neither increased nor stayed the same)

Days Enrolled in Treatment and Substance Use Reduction Success by Demographic Characteristics

Considering time spent enrolled in program treatment and substance use reduction success for 2015-2016 fiscal year clients by demogrpahic characteristics`","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":122,"value":"viewof showVar = Inputs.select([\n \"gender\",\n \"age_group\",\n \"race\",\n \"race_general\",\n \"disability\",\n \"start_substance\", \n], {label: \"View By:\", value: \"gender\"})","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":160,"value":"viewof xVar = Inputs.select([\n \"program_success\",\n \"enrollment_v_target\",\n], {label: \"X-axis Variable:\", value: \"program_success\"})","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":147,"value":"viewof showFacet = Inputs.select([\n true,\n false,\n \n], {label: \"Split By Days of Enrollment:\", value: false})","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":556,"value":"viewof normalize = Inputs.select([\n true,\n false,\n \n], {label: \"Normalize Bars to 100%:\", value: true})","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":112,"value":"$3f","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":454,"value":"html`

Client Characteristics By Cluster

A breakdown of client characteristics for each determined client cluster`","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":337,"value":"$40","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":170,"value":"$41","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":324,"value":"html`

Days Enrolled in Treatment and Substance Use Reduction Success by Cluster

Considering time spent enrolled in program treatment and substance use reduction success for 2015-2016 fiscal year clients by cluster`","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":212,"value":"$42","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":442,"value":"md`### 2.1 Interpreting the Clusters`","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":445,"value":"$43","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":433,"value":"html`

`","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":47,"value":"md`#### Data`","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":88,"value":"data_raw = await FileAttachment(\"data_raw.csv\").csv({typed: \"auto\"})","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"data_raw","database":"$undefined"},{"id":91,"value":"k_explore = await FileAttachment(\"k_explore.csv\").csv({typed: \"auto\"})","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"k_explore","database":"$undefined"},{"id":166,"value":"k_explore_tidy = await FileAttachment(\"k_explore_tidy.csv\").csv({typed: \"auto\"})","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"k_explore_tidy","database":"$undefined"},{"id":78,"value":"data_mean_race_month = d3.csv(\"https://gist.githubusercontent.com/carolinecullinan/e94ac14b48b99ae32fc423ee94261cec/raw/993ec1d1a964c6726ae0d4c200dd4a30360c19be/mean_race_month.csv\", d3.autoType)","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":68,"value":"data = d3.csv(\"//gist.githubusercontent.com/carolinecullinan/c98b1e58a4c4bc59d4b274de21a6bfa1/raw/00686802acac8917c9e272580005331111cb34fa/data.csv\", d3.autoType)","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":285,"value":"md`### Modules`","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":288,"value":"import {Plot, d3} from \"@observablehq/plot\"","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":330,"value":"import { tol } from '@jotasolano/paul-tol-schemes'","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":282,"value":"md`### Styling`","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","tool":"$undefined","providerMetadata":"$undefined","attachments":"$undefined","output":"$undefined","database":"$undefined"},{"id":271,"value":"html`\n