Rewriting a patient leaflet to plain language: SMOG, Dale–Chall and passive voice, before and after

A small clinic wants its medication handouts readable by every patient who picks one up, not just the ones who already know what "gastrointestinal disturbances" means. The usual instruction is to "write at a sixth-to-eighth-grade level". The American Medical Association's health literacy manual for clinicians recommends sixth grade for patient materials, and the National Institutes of Health guidance on easy-to-read health materials is usually cited as eighth grade or below; a 2021 review of patient materials in high-impact medical journals (PMC8205335 on PubMed Central) summarises the pair as sixth and eighth grade respectively. That gives a target band, but not a method.

This guide is the method, run on one paragraph. It uses three tools in a fixed order: the SMOG index to set the goal and measure progress, Dale–Chall to find the specific words to replace, and the passive voice detector to find the sentences that hide who has to do what. Every score below came from pasting the text into the tool.

The leaflet text in this guide is an illustrative example written for the exercise, not medical advice. Follow the instructions your pharmacist or prescriber gives you.

The paragraph we started with

This is the kind of text that turns up on real handouts, often copied from the professional product information:

"Amoxicillin is a penicillin-type antibiotic indicated for the treatment of bacterial infections. It is essential that the full course of medication prescribed by your physician is completed, even if symptoms have subsided before the medication is finished. Discontinuation of treatment prematurely may result in recurrence of the infection or the development of antibiotic resistance. Each dose should be administered at regular intervals, preferably every eight hours, and may be taken with or without food. If gastrointestinal disturbances such as nausea or diarrhea are experienced, administration with a meal is recommended. Should a dose be missed, it should be taken as soon as it is remembered, unless the next scheduled dose is imminent, in which case the missed dose should be omitted. Patients are advised not to take two doses simultaneously to compensate for a missed dose. Immediate medical attention should be sought if a rash, swelling of the face or lips, or difficulty breathing is experienced, as these may indicate a serious allergic reaction."

It is accurate and complete. It is also 164 words in 8 sentences, averaging 20.5 words a sentence and 1.77 syllables a word.

Step 1: SMOG sets the target

The SMOG Index Calculator is the usual first check for health text because McLaughlin built it around full comprehension rather than getting the gist. It counts polysyllables, meaning words of three or more syllables, and ignores everything else. The published method samples 30 sentences from a long document. This implementation instead counts polysyllables over whatever you paste and scales them to a 30-sentence equivalent:

SMOG = 1.043 × √(polysyllables × 30 ÷ sentences) + 3.1291

The original paragraph has 41 polysyllables in 8 sentences. Scaled up, that is 153.75; the square root is 12.40; times 1.043 is 12.93; plus 3.1291 gives SMOG 16.1, which the tool labels "Postgraduate". The target is 6 to 8. Ten grades is not a tidy-up. It is a rewrite, and the formula tells you exactly where the grades are: in those 41 long words, and in packing them into only eight sentences.

Note what that formula implies. Sentence count sits in the denominator, so splitting sentences lowers SMOG even when you keep every word. Polysyllables sit under a square root, so removing them has diminishing returns: the first twenty you cut do much more than the last five.

Step 2: first edit, structure only

A sensible first pass changes structure and leaves vocabulary alone, so you can see what each kind of edit is worth. Each long sentence was split, and every instruction was turned round to address the reader directly ("Complete the full course", "If you miss a dose, take it when you remember"). Specialist words such as "administer", "imminent" and "simultaneously" stayed.

Measure	Original	Structure edit	Plain-language edit
Words	164	137	131
Sentences	8	13	12
Words per sentence	20.5	10.5	10.9
Polysyllables	41	35	7
SMOG	16.1	12.5	7.5
Dale–Chall raw score	10.2	9.9	5.9
Unfamiliar words (Dale–Chall)	35.4%	36.5%	10.7%
Passive sentences	6 of 8 (75%)	1 of 13 (7.7%)	0 of 12
Flesch–Kincaid grade	13.3	9.5	3.2
Flesch Reading Ease	35.9	45.5	91.8

Halving sentence length took SMOG from 16.1 to 12.5 and Flesch–Kincaid from 13.3 to 9.5. That is a real gain, and it is also where most editing stops, still four grades above target. The structure edit barely moved Dale–Chall (10.2 to 9.9) and actually raised the share of unfamiliar words, because the short connecting words that were cut were all familiar ones. Vocabulary is the remaining problem, and the next tool is the one that names it.

Step 3: Dale–Chall turns the score into a word list

The Dale–Chall Readability Calculator checks each word against a list of about 3,000 words familiar to most fourth graders, after stripping common endings such as plural "-s", "-ed" and "-ing" (so "changed" and "cities" match "change" and "city"). Its most useful output is not the score but the list of words it flags. For the structure edit it flagged 50, including administer, imminent, omit, simultaneously, compensate, discontinuing, prematurely, recurrence, preferably, intervals, gastrointestinal, disturbances and physician. That list is the rewrite brief. Each one gets a plain swap or a rephrase:

• "administer each dose at regular intervals" became "space your doses evenly through the day"
• "imminent ... omit the missed dose" became "if your next dose is due soon, skip the one you missed"
• "simultaneously to compensate" became "at once to make up for"
• "gastrointestinal disturbances" became "if it upsets your stomach"
• "recurrence of the infection" became "the infection can come back"

The final version:

"Amoxicillin is an antibiotic. It kills the germs that cause some infections. Take every dose until the pack is empty, even if you feel better. If you stop too soon, the infection can come back. Space your doses evenly through the day, about eight hours apart. You can take it with or without food. If it upsets your stomach, take it with a meal. If you miss a dose, take it as soon as you remember. But if your next dose is due soon, skip the one you missed. Never take two doses at once to make up for a missed one. Get help right away if you get a rash, your face or lips swell, or you find it hard to breathe. These can be signs of a serious allergy."

Dale–Chall still flags 14 words in it: amoxicillin, antibiotic, germs, infection(s), dose(s), stomach, rash, serious and allergy. Most of those should stay. A patient needs the drug's name and the word "dose" printed on the label, and there is no plainer word for "rash". This is the right way to read the list: every flagged word is either replaced or kept on purpose. Also cut was the antibiotic-resistance clause, which the clinic agreed belongs in a separate leaflet rather than in dosing instructions.

One note on reading the Dale–Chall number. It is a raw score, not a grade level, and the tool shows it that way with its band beside it. On the published conversion table a raw 5.0 to 5.9 corresponds to grades 5 and 6, 7.0 to 7.9 to grades 9 and 10, 9.0 to 9.9 to college, and 10.0 or more to college graduate. The original's 10.2 is labeled "College graduate", the structure edit's 9.9 "College (grades 13–15)", and the final 5.9 "Grades 5–6", inside the clinic's target. The Dale–Chall explainer has the full table.

Step 4: the passive voice detector checks who does what

In patient instructions, passive voice is not a style point. "The full course of medication ... is completed" does not say who completes it. "Immediate medical attention should be sought" does not say by whom, or how. The Passive Voice Detector found passive constructions in 6 of the original's 8 sentences (75%), including a single sentence with four of them: "be missed", "be taken", "is remembered" and "be omitted".

The structure edit took that to 1 of 13. The survivor was "before the medication is finished", which is arguably not a true passive at all (it describes a state), and the final version dropped it anyway. The plain-language version scores 0 of 12.

The detector's rule is simple, and knowing it helps you judge its output. It looks for a form of "be" (is, are, was, be, been and so on) followed by something that looks like a past participle, allowing "not", an "-ly" adverb or another auxiliary in between: a word ending in "-ed", a longer word ending in "-en", or an irregular form from a built-in list. That catches "should be kept in the fridge" and "was prescribed for you". It misses passives built with "get", so "if the tablet gets crushed" passes. And it flags adjectives that look like participles, so "you are tired" counts as passive while "you may feel tired" does not. The post on how these detectors work goes further into the edge cases.

Where SMOG on a single paragraph stops being reliable

A final version at SMOG 7.5 meets the 6 to 8 target, but the number deserves the same scrutiny as the text. Three limits apply to this implementation.

Short text is scaled up, and so is its noise. The 30-sentence normalisation means a 12-sentence paragraph has its polysyllable count multiplied by 2.5 before the square root. Score smaller pieces of the same final text and the result moves: its first three, four or five sentences each score SMOG 8.8, against 7.5 for the whole paragraph, because the opening holds four of the paragraph's seven polysyllables (amoxicillin, antibiotic, infections, infection). For a full leaflet, paste the whole leaflet. If the clinic needs a figure that matches the published method, the text should run to at least 30 sentences.

The syllable counter is a heuristic. It counts vowel groups and applies a few rules for silent endings. It reads "serious" as two syllables and "diarrhea" as two, so some medical words that really are polysyllables slip through, and occasional ordinary words are overcounted. Before trusting a borderline score, look at which words it counted.

No formula checks meaning. The final version scores well partly because every sentence is short. A leaflet can hit SMOG 7 and still leave out the one thing a patient needed, or say "skip the one you missed" where the pharmacist wanted something more specific. The scores are a floor for readability, not a sign-off. Test the leaflet on a few real patients, and have a pharmacist check that the plain version still says what the original said.

The routine, for the next leaflet

1. Run SMOG on the whole leaflet and note the gap to grade 6 to 8.
2. Split long sentences and rewrite instructions in direct address. Re-run SMOG; expect a few grades, not the whole gap.
3. Run Dale–Chall and work through the flagged-word list, replacing or deliberately keeping each one. Re-run SMOG.
4. Run the passive voice detector and make sure every instruction names who acts, usually "you".
5. Optionally, check the Sentence Length & Variety Analyzer. Here the final version averages 10.9 words with a range of 4 to 21; very uniform short sentences read as choppy, so a little variation is fine.

On this paragraph the routine took SMOG from 16.1 to 7.5 and cut the word count by a fifth. The SMOG explainer covers why health writers prefer it to the other formulas. The work that moved the score most was the boring part: one flagged word at a time.