11 minutes
Who reaches the count Division numbers describe the people who found the door, got through it and were recorded correctly. Everything that happens before the record is a filter, and the filters do not operate at random.
What you’ll be able to do Identify the denominator behind a division number and name at least three groups the count cannot include, using one report you actually read. Compare the responses to this question and explain your choice: What is the most useful thing to raise? Document a next step for Who reaches the count: Take one report you read regularly, write the denominator of its headline number in a single sentence, and list three groups that cannot appear in it. A record of us, not a record of them Almost every number a division reports is made of records, and a record exists because something happened between a person and the department: a referral arrived, an application was filed, an assessment was entered, a service was billed, a case was closed with a code. That is a real and useful thing to count. It is not the same as counting the people the division exists for.
Think about everything that has to go right before a person can appear in a count. Someone had to tell them the service existed, in a language and a format they could use. They had to reach a door that was open at a time they could get to it. They had to get through a first conversation, sometimes with an interpreter who may or may not have been arranged. Someone had to enter them correctly, in this system rather than another one. Only then are they countable. Every one of those steps is a filter, and none of them filters people at random.
This is why a division can run an honest, careful report and still describe a population that does not exist. The report describes the people the process managed to hold on to. The question that turns it into something you can act on is short: what is this a share of, and who could never be in it?
Five places a person falls out of a count Nobody referred them A discharge planner, a school, a clinic or a county worker either knows about the service or does not. Referral patterns are relationships, and relationships are uneven across regions, hospital systems and languages. A person nobody referred has no record anywhere, and no report will ever show a space where they should have been.
The first contact did not work A phone line open on weekday mornings. A voicemail box that fills. A return call from a blocked number. An intake conversation held in English because the interpreter line was busy and the appointment could not be moved. Each of these ends with no record at all, or with a record coded as no response. The person tried. The count shows nothing, or shows them declining.
They were recorded somewhere else Two lead agencies enter the same support under different codes. A person moves between counties and starts a new record. A service is billed in a category the report filters out. The person exists in the data and is missing from this number, which is considerably harder to notice than being missing entirely.
They left before the measure started Most process measures start a clock at some event: a scheduled date, an approved plan, an opened case. Anyone who leaves before that event is not late, not on time, and not counted. A measure with a soft entry point will always look better than the process it describes, and it will look best exactly where the process is failing earliest.
The group was too small to publish Suppressing small cells protects people from being identified, and it is the right rule. Its side effect is that the smallest communities vanish from the chart, then from the discussion, then from the priority list. Suppression is a reporting decision, not evidence that nothing is happening.
The same measure, four denominators Of tracked cases Of applicants Of people found eligible Of people who could benefit
The most available and the narrowest. It answers one question: of the cases that entered our workflow with a start event, how many finished on time? Genuinely useful for managing the workflow. It says nothing whatever about who never entered it.
Wider, and usually obtainable. It answers: of everyone who asked, how many got through? The space between this figure and the one above is a count of the people the workflow lost early, and that space is often exactly where the access problem lives.
This one asks a fairness question. Among people the program has already agreed it is for, is the experience even? A difference here is difficult to explain away, because eligibility has already been settled and the remaining variation belongs to the process.
The question leaders usually mean, and the one administrative records cannot answer. Getting close to it takes population estimates, community knowledge, and compensated partners who can say who is not showing up and why. Naming the distance between this question and your available numbers is far more honest than quietly answering a smaller question instead.
Ask for the denominator before you ask for the trend What you can change You control which question you ask first when a figure arrives in front of you, and whether the sentence explaining what it counts travels with it.
What to watch for Do not accept a rate without its denominator, and do not let a target be set on a measure with a soft entry point. Both are how a process becomes better at reporting rather than better at serving people.
Your next step The next time a rate is presented to you, ask in the room what it is a share of, and what happens to a case that leaves before the count starts. Ask it as curiosity rather than challenge. Most of the time nobody has been asked.
Where does this number lose people? A person whose hospital discharge planner has never heard of the service.
Choose a category Lost before any record exists Lost inside the record Lost in how it is reported A preferred-language field left blank in a third of records.
Choose a category Lost before any record exists Lost inside the record Lost in how it is reported A group of six people suppressed because the cell is too small to publish.
Choose a category Lost before any record exists Lost inside the record Lost in how it is reported Someone who called the general line three times, never reached a worker, and stopped calling.
Choose a category Lost before any record exists Lost inside the record Lost in how it is reported Two lead agencies entering the same support under different codes.
Choose a category Lost before any record exists Lost inside the record Lost in how it is reported A chart that shows only the four largest regions.
Choose a category Lost before any record exists Lost inside the record Lost in how it is reported A case closed as no response after an intake call that had no interpreter.
Choose a category Lost before any record exists Lost inside the record Lost in how it is reported A statewide total that leaves out anyone whose case is still open.
Choose a category Lost before any record exists Lost inside the record Lost in how it is reported Check matches Reset I ran that timeliness number for three years and I was proud of it. Then a county supervisor asked me what happened to the referrals that never got a scheduled date. So I went and looked. There were more of them than I expected, they were concentrated in two regions, and not one of them had ever appeared in anything I sent upward. The number was not wrong. It was just about a smaller group of people than everyone reading it believed.
Composite DSD staff perspective, illustrative A private reflection, for you only and never collected: the number I am asked about most often is... what it is actually a share of is... and the people who cannot appear in it are... If the second line takes more than one sentence, that is worth knowing before the figure is used again.
Carry this forward Administrative numbers are a record of transactions with the division, not a census of the people the division exists for. Anyone who was never referred, never got through a first contact, or was recorded somewhere else is outside the count and invisible inside it.
The denominator decides the question. Share of tracked cases, share of applicants, share of people found eligible and share of people who could benefit are four different measures, and usually only the first two are available.
Undercounting is patterned, not random. It concentrates among people who needed an interpreter and did not get one, people without a stable address, people in counties with a single access point, and people nobody referred.
Absence reads as satisfaction unless you name it. No complaints from a language community is a finding about the complaint route, not a finding about the experience.
A scenario about a timeliness figure with a hidden denominator, a sorting exercise that places losses before, inside and after the record, and a knowledge check on what a completion rate can support.
Take one report you read regularly, write the denominator of its headline number in a single sentence, and list three groups that cannot appear in it.
Browse and download only. Course notes are not typed or saved on this page.
Reset this lesson
Course overview Next lesson Participation and course completion in this program do not count toward DHS-required training credits unless management, a director, or DHS leadership expressly approves an exception.
